Hochschule für Angewandte Wissenschaften Hamburg

University of Applied Sciences

Faculty of Life Sciences

M.Sc. Health Sciences

Real-world evaluation of adverse pregnancy outcomes in women with

gestational diabetes mellitus in the German health care system

-

A health claims data analysis

Master Thesis

Date of Submission: 25.09.2018

Submitted by: Patrick Reinders

Matriculation number:

Examination supervisor: Prof. Dr. York Francis Zöllner

Secondary supervisor: P.D. Dr. Udo Schneider

I

Table of contents

1. Introduction ................................................................................................................... 1

2. Research question .......................................................................................................... 3

3. Gestational diabetes mellitus ......................................................................................... 4

3.1. Pathophysiology ...................................................................................................... 4

3.2. Epidemiology .......................................................................................................... 5

3.3. Common maternal and foetal adverse pregnancy events ........................................ 7

3.4. National and international recommendations in screening and treatment of the

gestational diabetes mellitus .............................................................................................. 9

3.5. Health care service research reports in Germany ................................................. 14

4. Methods ....................................................................................................................... 18

4.1. Health care service research with SHI - Health claims data ................................. 18

4.2. Identification of study groups ............................................................................... 19

4.3. Periods of the analysis and definition of outcome parameters ............................. 20

4.3.1. Period of pregnancy "t0" ................................................................................... 21

4.3.2. Period of delivery "t1" ....................................................................................... 23

4.3.3. Period of follow-up "t2" .................................................................................... 25

4.4. Linkage of mother and child ................................................................................. 25

4.5. Validation of the "obstetric comorbidity index" in a German setting .................. 27

5. Results ......................................................................................................................... 30

5.1. Period of pregnancy “t0” ...................................................................................... 32

5.2. Period of delivery “t1” .......................................................................................... 43

5.3. Period of follow-up “t2” ....................................................................................... 51

6. Discussion .................................................................................................................... 54

6.1. Comparing of findings with current research ....................................................... 54

6.2. Advantageous of the study .................................................................................... 57

6.3. Limitations of the study ........................................................................................ 59

7. Conclusion ................................................................................................................... 61

II

8. Excurse: Validation of the obstetric comorbidity index .............................................. 62

8.1. Results of the validation ....................................................................................... 62

8.2. Discussion of the validation .................................................................................. 65

9. References ................................................................................................................... 67

10. Appendix .................................................................................................................. 77

10.1. Methods ............................................................................................................. 77

10.2. Results ............................................................................................................... 81

III

Table of tables

Table 1 Blood glucose targets Source: (Kleinwechter et al., 2011, p. 31) .......................... 12

Table 2 Doctor specialty groups .......................................................................................... 23

Table 3 Operationalization of HAPO outcomes .................................................................. 24

Table 4 Time of test execution, 50g GCT ........................................................................... 33

Table 5 Time of test execution, 75g OGTT......................................................................... 33

Table 6 Utilization of 50g GCT by doctor specialist groups ............................................... 35

Table 7 Utilization of 75g OGTT by doctor specialist groups ............................................ 35

Table 8 Cross-tabulation: Blood glucose self-monitoring * screening ............................... 36

Table 9 Doctor specialty group prescribing blood glucose stripes ...................................... 37

Table 10 Diagnoses coded by doctors ................................................................................. 39

Table 11 Cross-table diagnoses by specialist groups: Gynaecology * diabetology ............ 40

Table 12 Cross tabulation: Blood glucose stripes * GDM diagnosis .................................. 40

Table 13 Initial prescriptions of insulin by specialist groups .............................................. 41

Table 14 Cross-table Insulin * Diagnosis by a diabetologist .............................................. 42

Table 15 Cross-tabulation: Insulin * blood glucose stripes ................................................. 42

Table 16 Baseline characteristics ........................................................................................ 44

Table 17 General characteristics of the newborn ................................................................ 45

Table 18 Frequency of the HAPO Outcomes by groups ..................................................... 46

Table 19 Logistic regression: Birth weight > 90th percentile .............................................. 47

Table 20 Logistic regression: Primary caesarean section .................................................... 48

Table 21 Logistic regression: Neonatal hypoglycaemia...................................................... 49

Table 22 Odds ratios for primary and secondary outcomes: Group B ................................ 50

Table 23 Odds ratios for primary and secondary outcomes: Group C ................................ 51

Table 24 Follow-up 75g OGTT ........................................................................................... 52

Table 25 75g OGTT by doctor specialty groups ................................................................. 53

Table 26 Distribution of OCI and end-organ damage ......................................................... 62

Table 27 Distribution of the Elixhauser Score and end-organ damage ............................... 63

Table 28 Frequency of comorbidities of the obstetric comorbidity index .......................... 64

Table 29 ICD-10 codes used for identification of pregnancies ........................................... 77

Table 30 OPS-codes used to identify pregnancies .............................................................. 78

Table 31 ICD-10 codes defining comorbidities within the OCI ......................................... 79

Table 32 ICD-10 codes defining end-organ damage ........................................................... 80

IV

Table 33 Logistic regression Group B: Birth weight >90th percentile ............................... 81

Table 34 Logistic regression Group B: Primary cesearan sectio ........................................ 82

Table 35 Logistic regression Group B: Neonatal hypoglycaemia ....................................... 83

Table 36 Logistic regression Group B: Premature delivery ................................................ 84

Table 37 Logistic regression Group B: Shoulder dystocia .................................................. 85

Table 38 Logistic regression Group B: Intensive Neonatal care ......................................... 86

Table 39 Logistic regression Group B: neonatal hyperbilirubinemia ................................. 87

Table 40 Logistic regression Group B: Preeclampsia ......................................................... 88

Table 41 Logistic regression Group C: Premature delivery ................................................ 89

Table 42 Logistic regression Group C: Shoulder dystocia .................................................. 90

Table 43 Logistic regression Group C: Intensive neonatal Care ......................................... 91

Table 44 Logistic regression Group C: Hyperbilirubinemia ............................................... 92

Table 45 Logistic regression Group C: Preeclampsia ......................................................... 93

Table of figures

Figure 1 Periods of the analysis ........................................................................................... 21

Figure 2 Linkage of mothers to their baby .......................................................................... 26

Figure 3 Receiver operator characteristic curve by (Zou et al., 2007) ................................ 28

Figure 4 Flowchart Study population: data year 2016 ........................................................ 31

Figure 5 Screening test in pregnant women ........................................................................ 32

Figure 6 Week of pregnancy, when tests were applied ....................................................... 34

Figure 7 Administrative prevalence with different degrees of validity ............................... 38

Figure 8 Frequency of primary outcomes across study groups ........................................... 45

Figure 9 Cumulative 1-year rate of diabetes mellitus type 2 ............................................... 53

Figure 10 ROC curve: OCI and Elixhauser Score ............................................................... 65

V

List of abbreviations

ACOG American College of Obstetricians and Gynecologists

ADA American Diabetes Association

ATC Anatomical therapeutic chemical classification system

AUC Area under the curve

ACHOIS Australian Carbohydrate Intolerance Study in Pregnant Women trial

BMI Body mass index

DDD Defined daily dose

DDG Deutsche Diabetes Gesellschaft (German Diabetes Association)

DGGG Deutsche Gesellschaft für Gynäkologie und Geburtshilfe (German

Society Of Obstetrics And Gynaecology)

DMT1 Diabetes mellitus type 1

DMT2 Diabetes mellitus type 2

DMP Disease Management Program

EBM Einheitlicher Bewertungsmaßstab (doctor's fee scale)

FPG Fasting plasma glucose

G-BA Gemeinsamer Bundesauschuss (Federal joint committee)

GP General practitioner

G-DRG German diagnosis related groups

GDM Gestational diabetes mellitus

GCT Glucose challenge test

HAPO Hyperglycaemia and Adverse Pregnancy Outcomes

IADPSG International Association of Diabetes and Pregnancy Study Groups

ICD-10-GM International Classification of Diseases 10th revision, German

modification

KBV Kassenärztliche Bundesvereinigung (National Association of

Statutory Health Insurance Physicians)

LANR Lebenslange Arztnummer

MFMU Maternal Foetal Medicine Units Network trial

NICE National institute for health and care excellence

OCI Obstetric comorbidity index

OPS Operationen- und Prozedurenschlüssel (German operation and

procedure codes)

VI

OGTT Oral glucose tolerance test

ROC Receiver operator characteristic curve

SHI Statutory health insurance

TK Techniker Krankenkasse

WHO World Health Organization

1

1. Introduction

Gestation is a stressful period for the female body, while metabolic adaptions appear to

nurture the foetus. For some women, these adaptions result in hyperglycaemia, also known

as gestational diabetes mellitus (GDM) (Catalano, Huston, Amini, & Kalhan, 1999). The

condition is linked to short and long-term health burden for mother and child. In the short-

term, GDM is associated with adverse pregnancy events, such as shoulder dystocia, neonatal

hypoglycaemia, neonatal hyperbilirubinemia, and maternal preeclampsia (Metzger et al.,

2008). In the long-term, mother and child have a higher risk to develop diabetes mellitus

(Clausen et al., 2008; Song et al., 2018).

At the beginning of classifying the condition, only pregnant women with a risk profile were

tested for the disease. The risk profile included age, obesity, and women, whose close

relatives got diagnosed with diabetes mellitus. Aim of this strategy was to reduce the risk for

mothers to become diabetic in the years after pregnancy. The approach changed

fundamentally, when the mild form of GDM was linked to the described adverse pregnancy

events by the Hyperglycaemia and Adverse Pregnancy Outcome Study (HAPO) (Metzger et

al., 2008). In 2010, the International Association of Diabetes and Pregnancy Study Groups

(IADPSG) published new criteria and recommended a new screening strategy, based on the

result of the HAPO study. The threshold to be diagnosed with the condition is lower than

the previous threshold in order to diagnose women with a milder form of hypoglycaemia.

Furthermore, the IADPSG recommends testing all women, regardless of a risk profile

(IADPSG, 2010). The new established IADPSG-criteria have been adopted internationally

by multiple guidelines (ADA, 2017; Kleinwechter et al., 2011; NICE, 2015).

In Germany, the federal joint committee (Gemeinsamer Bundesausschuss; G-BA) adopted

the criteria in 2011 and implemented a population wide screening algorithm into the German

health care setting. However, instead of a one phase screening, recommended by the

IADPSG, a two-phase approach has been established in Germany (G-BA, 2012). In addition

to the screening, a guideline was developed by the German Diabetes Association (Deutsche

Diabetes Gesellschaft; DDG) in cooperation with the German Society Of Obstetrics And

Gynaecology (Deutsche Gesellschaft für Gynäkologie und Geburtshilfe; DGGG). The main

objective of the guidelines is to reduce the risk of adverse pregnancy events. Early

identification and treatment with behavioural and pharmacological intervention via insulin

were included measures to reduce the burden of the disease. Even though an impact

measurement of the guideline in a two-year schedule was recommended by the authors of

2

the guideline, population wide data analysing the health of mother and child is missing

(Kleinwechter et al., 2011). Epidemiological studies, analysing the current care situation of

pregnant women with GDM and their pregnancy outcomes, are necessary to be able to

compare data over time, with other countries, and other treatment frameworks.

Only one study, using outpatient data, analysed the complication rate and the impact of the

new implemented screening schedule on the health of the mother in Germany. However,

described neonatal complications were not studied, while linked mother-baby pairs could

not be established with the present database. As the health of the newborn is crucial and was

one of the main reasons for adapting the classification of the disease, data on complication

rates of neonatal specific outcome is needed (Tamayo, Tamayo, Rathmann, & Potthoff,

2016).

Therefore, this master thesis used data from a large statutory health insurance (SHI) to assess

the health burden and utilization of services, recommended within the guideline, of pregnant

women diagnosed with GDM. SHI data has the advantage that all services and diagnoses

from the in- and outpatient sector are available (Ohlmeier et al., 2014). Furthermore, it is

possible to link mothers to their child and analyse neonatal adverse pregnancy events (Garbe,

Suling, Kloss, Lindemann, & Schmid, 2011).

The research questions are presented in chapter 2. This is followed by essential background

information of the GDM in chapter 3, including pathophysiology, epidemiology, description

of adverse pregnancy events, national and international treatment recommendations, and

current health care service research reports. Knowledge about German health claims data

and methodological considerations to fulfil the objectives are presented in chapter 4. The

results of the study, with focus on the rate of adverse pregnancy events are accessible in

chapter 5. In the next chapter (chapter 6) a discussion is subjected, comparing the results to

other research as well as examining the strength and limitations of the study. Finally, a

conclusion of the report will be stated in chapter 7.

,

3

2. Research question

The S3-guideline, which introduced international accepted diagnostic criteria and treatment

recommendations for the GDM into the German health care system, was established to lower

the burden of GDM. However, it remains unclear to which degree the guideline has been

adopted and further how the new diagnostic criteria and treatment recommendations

impacted the health of mother and child. Therefore, one primary and two secondary

objectives are examined in the master thesis. The primary objective is: to compare the

occurrence of adverse clinical outcomes between obstetric women with GDM to women

without a GDM. In order, to assess the quality of results of different treatment escalation

levels within the guideline, the GDM group will be divided into a group receiving

pharmacological treatment and a group not. The secondary objectives are: Firstly, to evaluate

the application of the screening process and the referring to a specialist, if the diagnosis was

determined. Secondly, to assess the proportion of women utilizing follow-up care for the

GDM after delivery.

4

3. Gestational diabetes mellitus

The following chapter provides a general overview of important background information of

the disease. This includes pathophysiology, epidemiology, adverse events, screening, and

treatment options. The focus is on the current screening and treatment options recommended

by the German guidelines of 2011, as those have an impact on the outcome of the study. The

provided background information will help to understand the chosen research question, the

disease itself, but also the status of research and problematic discussion points in the

provision of care for these women.

3.1. Pathophysiology

Pregnancy is a stressful period for the metabolism of the female body. The metabolism

adapts in order to nurture the foetus from the beginning of pregnancy until birth.

Physiological adaption occurs especially during late pregnancy, due to foetal growth and an

increased need for calories of the foetus. Research observed that carbohydrates are preserved

in the body of the mother, as they are an important fuel for mother and child. To preserve

those, an resistance of insulin starts to develop during mid-pregnancy, with progression over

time (Sivan, Homko, Chen, Reece, & Boden, 1999). insulin resistance increases by 50-60%

within normal female bodies (Catalano, 2014). This adaption is largely initiated by placental

hormones, which also explains why the resistance disappears after pregnancy. As the

resistance intensifies over time, so does the insulin response to compensate for the resistance.

In female bodies with a normal glucose regulation during the obstetric period, glucose levels

remain nearly consistent (Buchanan & Xiang, 2005). However, when the high demand of

insulin in the time of pregnancy cannot be met, it leads to hyperglycaemia (Catalano, 2014).

The mismatch between supply and demand of insulin is explained by the dysfunction of

pancreatic-β-cell that is apparent in women with GDM and is responsible for the low

secretion rate of insulin during late pregnancy (Homko, Sivan, Chen, Reece, & Boden,

2001). The insufficient insulin response of the pancreas in pregnant the body is unmasked

by the increasing insulin resistance in late pregnancy. Hyperglycaemia with first onset in

pregnancy, particularly in late pregnancy, is then defined as the gestational diabetes mellitus

(Buchanan & Xiang, 2005).

5

Postpartum, in women with normal glucose tolerance, the pregnancy related adaption of the

metabolism disappears. Insulin resistance and secretion usually have the same level as they

had before the pregnancy. However, in women with a GDM, those observations cannot be

made. Women with the condition have an impaired insulin resistance and secretion after

pregnancy. GDM postpartum is therefore considered as a prediabetic state and may lead

sooner or later to the manifestation of a type 2 diabetes mellitus (Kautzky-Willer et al.,

1997).

While pregnant, constant hyperglycaemia is thought to influence the metabolism of the

foetus and leads to foetal overgrowth or macrosomia. The Pedersen hypothesis or the

hyperglycaemia-hyperinsulinemia hypothesis first tried to explain the phenomenon. Hereby,

the mother’s hyperglycaemia leads to a foetal hyperglycaemia and finally hyperinsulinemia

or an increased foetal insulin response. Both factors, the increased amount of calories for the

foetus and hyperinsulinemia, results in exaggerated fat stores and hence macrosomia

(Macfarlane & Tsakalakos, 1988). However, research is available to disprove the hypothesis

in parts. A study with the aim to identify the relationship of maternal glucose and lipids on

macrosomia, came to the conclusion that the mother’s lipid levels had a higher influence on

foetal growth than maternal glucose levels (Schaefer-Graf et al., 2008). A different report of

the HAPO study, focusing on the relationship between foetal complications, with focus on

foetal growth, identified a strong association between the mother body mass index and

macrosomia, even when adjusted for maternal glucose levels (McIntyre et al., 2010).

3.2. Epidemiology

Epidemiological figures for the GDM, especially the prevalence and the transition rate from

GDM to type 2 diabetes mellitus, are reported inconsistent. All estimates differ by the

definition of GDM, the screening algorithm, and the population under research (Eades,

Cameron, & Evans, 2017). In the beginning of establishing the diagnosis of GDM only

women with risk factors were tested with a glucose test. However, this changed widely after

a consensus meeting of the IADPSG in 2010. They not only proposed general screening for

all pregnant women but adapted common criteria as well (IADPSG, 2010).

In Germany, general screening and the new IADPSG criteria were implemented in the year

of 2011 in order to identify women with a hyperglycaemia during pregnancy (Kleinwechter

6

et al., 2011, p. 11). The latest prevalence estimate concluded a percentage of 13.2% for all

pregnant women in Germany, after the introduction of general screening. However, this

estimate is based on health-claims data and, therefore, many limitations apply, e.g. a not

verifiable validity of the diagnosis (Melchior, Kurch-Bek, & Mund, 2017).

A meta-analysis of the year 2017, with a goal to define the prevalence of GDM in Europe,

considered 40 studies from all over the world. The mean prevalence throughout all studies

was 5.4% [95% CI 3.8 – 7.8]. The prevalence was largely affected by the year of data

collection, diagnostic criteria, and the country where the study was conducted. For the latest

definition of GDM, the IADPSG criteria, a prevalence of 14.1% was estimated. For the last

period of data collection, between 2010 and 2016, a prevalence of 11.1% was calculated. In

comparison, the prevalence for the time period form 2000 – 2009 was only 6.9% (Eades et

al., 2017).

Criteria are constantly changing. Therefore, an increase of the prevalence due to lifestyle

factors, e.g. obesity, is hard to measure. Only a minor quantity of epidemiological studies

was able to detect such an effect. In Sweden, a ten years trend of the GDM prevalence was

observed. Over these ten years, the screening algorithm and diagnostic criteria remained

unchanged. Still, the prevalence increased from 1.9% [95% CI 1.8 – 2.0] in 2003 to 2.6%

[95% CI 2.4 – 2.7] in 2012. A clear reason for the development could not be discovered, but

it was suspected that risk factors increased over time (Ignell, Claesson, Anderberg, &

Berntorp, 2014). Similar developments were noticed in different populations and ethnicities

(Dabelea et al., 2005; Hunt & Schuller, 2007).

The most common reported factors that increase the risk of a GDM are advanced age,

increased weight, a family history of diabetes, parity, and a previous delivery of a large infant

(Ben-Haroush, Yogev, & Hod, 2004). Noteworthy, epidemiological studies focusing in these

risk factors are missing. International societies, however, still accept those as the main

factors for developing a GDM (ADA, 2017, p. 18; Zhang & Ning, 2011).

A recent epidemiological study analysed the main risk factors, age, and body mass index

(BMI) in women with multiple pregnancies. Women with a BMI above 30 in comparison to

those with a BMI under 30, had a significantly increased risk of developing a GDM (Odds

Ratio (OR) 4.88; p < 0.001). Also, older women, with an age above 35 years, were diagnosed

with GDM nearly two times more often in comparison to women below 35 years (OR 1.81;

7

p < 0.01) (Cozzolino et al., 2017). Even though, this was found in women with multiple

gestation, these risk factors will probably apply for women with single deliveries, as well.

At the beginning of defining GDM, the main goal was to prevent the establishment of

diabetes mellitus following a gestation. The reason for a treatment changed over time,

however, the transition rate to a diabetes type 1 or 2 is still a relevant epidemiological key

figure. In a recent meta-analysis, gathering data from 30 cohort studies with more than 2,5

million women, the long-term risk of developing diabetes was evaluated. For the short-term

outcome, 3 years or less, combined data from studies revealed a relative risk (RR) of 4.82

[95% CI 2.19 - 10.62] to become diabetic. When looking at the mid-term outcome, 3 - 6

years, the RR was 16.16 [95% CI 9.96 - 26.24] and therefore substantially increased. The

risk was largely different by age groups, BMI before/after pregnancy, and at follow up and

region. Women in the region of Europe, North America, and the Middle East had a high

associated risk to develop diabetes after GDM. Interestingly, no linear relationship between

an increased BMI or age and an increased risk of diabetes following a GDM was observed.

Women with age at follow-up of above 40 years, had a lower adjusted chance of having a

diabetes than women below the year of 35. Women with a BMI at follow up below 25, had

a higher chance to develop a diabetes than women with a BMI above (Song et al., 2018).

Currently, conclusive evidence explaining these findings is missing.

Not only mothers with GDM have an increased risk for the later development of diabetes,

their children do as well. Due to the long follow-up, only a few studies exist to investigate

this question. An observational study in Denmark detected an increased risk for diabetes

type 2 for the descendants of mothers with GDM. The adjusted odds ratio of being diagnosed

with the disease for these children was 7.76 [95% CI 2.58 – 23.39] (Clausen et al., 2008).

Apart from long-term effects in children born by mothers with hypoglycaemia, also short-

term events can occur. This will be described in the next chapter.

3.3. Common maternal and foetal adverse pregnancy events

The gestational diabetes mellitus is related to many maternal and foetal complications. As

this analysis of health claims data includes adverse events that were prior used in the HAPO

study, an overview of those and the connection to the GDM will be given. On the maternal

side the HAPO study included preeclampsia and primary caesarean section, while the foetal

8

side, neonatal hypoglycaemia, macrosomia, premature delivery, shoulder dystocia/birth

injury, and hyperbilirubinemia were included (Metzger et al., 2008).

Preeclampsia is a renal disease and evolves during late pregnancy. Hypertension,

proteinuria, and oedemas are the main symptoms of the disease. Untreated, it can evolve into

an eclampsia, which additionally includes severe seizures of the mother. Eclampsia is a

leading cause of maternal and foetal death (Al-Jameil, 2013; Sibai, Dekker, & Kupferminc,

2005). The pathogenesis and the relationship to gestational diabetes are not yet well defined.

However, a clear correlation between these two diseases can be found within the HAPO

study. An increased level of hypoglycaemia leads to an increased rate of preeclampsia

(Metzger et al., 2008).

Primary caesarean section and shoulder dystocia/birth injury are strongly related to

macrosomia of the foetus. In the list of relative and absolute indications for a section,

published by the DGGG, macrosomia is defined as relative indication for a primary section

(DGGG, 2010, p. 3). Even though there is no clear clinical evidence supporting the caesarean

section, surgical delivery is considered by experts at an increase birth weight above >5000g

(ACOG, 2016). It should be considered that, generally, caesarean sections are related to

negative long-term effects, including asthma and obesity for the new born and the risk of

future maternal pregnancy complications (Keag, Norman, & Stock, 2018). The reduction of

macrosomia could therefore lead to a reduction in the caesarean section rate. However,

research showed that even mothers with treated GDM, resulting in lower rates of

macrosomia, did not have lower rates of a caesarean delivery. The author concluded a

“labelling effect”. Gynaecologists have a lower threshold to conduct a caesarean delivery in

patients diagnosed with GDM in comparison to women were the birth weight of the child is

comparable (Naylor, Sermer, Chen, & Sykora, 1996).

A foetal outcome associated with increased birth weight, hence macrosomia, is shoulder

dystocia, a failure in the delivery of the shoulders of the foetus. Other risk factors for this

outcome are the maternal weight, abnormal labour, and previous shoulder dystocia.

Complications of the dystocia are fractures, brachial plexus palsy, and in rare cases even

death, with a low rate between 0.35% and 2.9% of all cases with shoulder dystocia. The

adverse event occurs only in vaginal deliveries with a rate between 0.3% and 3.0% (Gherman

et al., 2006).

9

Another primary outcome of the HAPO study is the clinical neonatal hypoglycaemia, a low

blood glucose level appearing in 48 hours after delivery. The energy deficit due to low blood

glucose levels results in seizures, coma, and long-term negative neurological outcomes, if

the hypoglycaemia persists for several hours (Kerstjens, Bocca-Tjeertes, de Winter,

Reijneveld, & Bos, 2012; Rozance & Hay, 2010). The HAPO study provided evidence that

the clinical neonatal hypoglycaemia is nearly linear associated with the mothers glucose

level (Metzger et al., 2008).

The increased understanding of the relationship between adverse pregnancy events and the

gestational diabetes mellitus led to a new definition of the disease and increased focus on

treatment to prevent those.

3.4. National and international recommendations in screening and treatment of the

gestational diabetes mellitus

In this section recommendations, reviews, and guidelines focusing on diagnostic criteria,

screening, treatment, and follow-up of the gestational diabetes mellitus will be displayed. As

this analysis observed the German health care reality, a special focus will be given to the

German guidelines of diagnostic, screening and treatment of the GDM from 2011

(Kleinwechter et al., 2011). The German guideline was recently updated in march 2018

(DGG & DGGG, 2018). Differences between these two versions will be shortly presented,

but the larger focus is on the guidelines from 2011, because this analysis observed a

population and their treatment from 2015 – 2017, so under the guidance of the older version.

3.4.1. Definition

The current definition of GDM has been established by the IADPSG in 2010 and was

adopted by guidelines all over the world, including Germany, the US, and Great Britain

(ADA, 2017; Kleinwechter et al., 2011; NICE, 2015). Hereby, all women are classified with

the impairment, if one of the three following criterions after a 75g oral glucose tolerance test

(OGTT) is equalled or exceeded: 1) Fasting plasma glucose (FPG) > 92mg/dl, 2) 1 hour

plasma glucose >180mg/dl, 3) 2 hour plasma glucose >153mg/dl (IADPSG, 2010).

10

As mentioned in chapter 2.2, the new criteria effected the prevalence of the GDM. Before

the establishment of the new criteria and the suggestion of a general population wide

screening, a risk assessment was recommended. No screening was required for women with

a low risk of GDM. The main goal of this approach was to reduce the increased risk of

mothers to become a diabetic after pregnancy. The new criteria were also established to

identify hyperglycaemic disorders of the mother, in order to prevent maternal and foetal

adverse events (IADPSG, 2010; Metzger & Coustan, 1998).

3.4.2. Screening

In 2011, the proposed cut-off values for the 75g OGTT were implemented into the German

health care system. The G-BA, the decision-making body of the self-administration in

Germany, decided on a two-step approach. Before the application of the OGTT, a 50g oral

glucose challenge test (GCT) for all pregnant women between the 24th and 27th week of

pregnancy is described. Only, if this test exceeds blood glucose levels of 135 mg/dl, the

second stage, namely the 75g OGTT, of screening may be applied. If the test exceeds levels

of >200 mg/dl the 75g OGTT is not required, and the GDM can be diagnosed right away.

From a reimbursement perspective, the OGTT test can only be remunerated, if the GCT test

has a positive result (G-BA, 2012, p. 7).

In the update of the S3-guideline for gestational diabetes mellitus, the administration of the

50g test is not recommended. Doctors should primarily offer the 75g OGTT, while there is

no evidence supporting the thresholds of the 50g GCT (DGG & DGGG, 2018). However, in

the guidelines of 2011 the full screening algorithm was recommended (Kleinwechter et al.,

2011).

IADSPG criteria were largely based on the HAPO study. HAPO was designed to identify a

threshold of the plasma glucose level, where the risk for adverse events substantially

increases. However, no clear threshold was identified in the study, therefore, a cut-off has

been defined by a consensus. The glucose level for the fasting plasma glucose, 1 hour and

2- hour 75g OGTT, were based on an increase of the ORs of 1.75 for the outcomes of birth

weight >90th percentile, cord C-peptide >90th percentile, and percent body fat >90th

percentile. The new criteria are controversial discussed. The previous chapter, which

described the epidemiology, already worked out that the IADPSG criteria increased the

prevalence of GDM, hence including women with a mild form of the impairment. On the

11

one hand, the diagnosis of a mild GDM could lead to enlarged medicalization of pregnant

women without a clear effect (Benhalima et al., 2016). On the other hand, in a Spanish cohort

study, comparing IADSPG criteria to the previous recommended Carpenter Coustan criteria,

a reduction in pregnancy adverse events was identified. The previous criterion, has a higher

threshold, hence leading to a lower prevalence of women diagnosed. The introduction of the

new criteria was even considered as cost-effective (Duran et al., 2014). The reason to reduce

the threshold of the 75g OGTT to included also a milder form of a GDM were two clinical

trials. Both trials focused on the treatment of a milder form of the disease and showed

positive effects on maternal and foetus adverse events (Crowther et al., 2005; IADPSG,

2010; Landon et al., 2009).

Even though the cut-off of the 75g OGTT is based on a consensus, the cut-off is validated

by clinical outcome parameters. There is no such validation or a consensus for the 50g GCT

test available. Within the German guideline, it is the main argument for not supporting the

50g GCT (DGG & DGGG, 2018, p. 20). However, a recent publication of a team of Belgium

scientists evaluated different thresholds of GCT tests as pre-test for the 75g OGTT test. At

the threshold chosen for the German screening guideline, >135mg/dl, a sensitivity of 66.2%

[95% CI 59.7 – 72.3] and specificity of 76.1% [95% CI 73.9 – 78.1] was identified. The

diagnostic test statistic of the GCT was described as moderate (Benhalima et al., 2018). The

article could further support a one-step approach instead of the current two-step approach.

The low sensitivity rate excludes more than 30% of women with a gestational diabetes

mellitus from a treatment. That is thought to reduce the rate of adverse events. Additionally,

around 25% of women with a positive test worry unnecessarily to be diagnosed with a

gestational diabetes mellitus before being tested with the 75g OGTT.

No sensitivity or specificity of the test can be calculated for the 75g OGTT. This test is

already the standard test to confirm the diagnosis and can therefore not be compared to a

gold standard. However, two problems occur in the administration of the test. One is a low

to medium reproducibility that might be more common in a pregnant population due to a

constant adaption in the metabolism. Two studies researching the reproducibility of a 100g

OGTT test came to an overlap of 78% (50 from 64 women tested) and 76% (29 of 38 women

tested) in the pregnant populations, measured in two consecutive weeks, respectively

(Catalano, Avallone, Drago, & Amini, 1993; Harlass, Brady, & Read, 1991). Another

problem is the fact that the test is challenging for pregnant women. First, it is time consuming

and, secondly, many women report nausea and vomiting when doing the test. Therefore,

12

other methods are proposed and will be challenged in the next years (Huhn et al., 2016;

Ryser Rüetschi et al., 2016).

3.4.3. Treatment of the GDM

After being diagnosed with GDM, all women should receive medical counselling by

diabetologist (Kleinwechter et al., 2011, p. 39). The main goal, when treating GDM, is to

normalize or keep blood glucose levels between specified blood glucose target levels. In the

German S3 guideline of 2011 and 2018 target levels (see Table 1) were aligned with two

clinical trials. The Australian Carbohydrate Intolerance Study in Pregnant Women

(ACHOIS) and the Maternal-Foetal Medicine Units Network (MFMU) trial (Crowther et al.,

2005; Landon et al., 2009). Measurement are elevated by self-monitoring.

Table 1 Blood glucose targets Source: (Kleinwechter et al., 2011, p. 31)

Time mg/dl blood-glucose plasma

Fasting, preprandial 65 - 95

1h postprandial <140

2h postprandial <120

Mean blood glucose level 90 – 110

Mean blood glucose level 80 – 100

All women should receive nutritional counselling in any way. The decision for a further

pharmacological therapy is indicated, if blood glucose levels are not normalized by lifestyle

interventions alone. Insulin therapy should be first initialized in the first two weeks after the

diagnosis of GDM, to wait for the effect of life style factors. Afterwards, the need for insulin

should be continuously evaluated (Kleinwechter et al., 2011).

Lifestyle interventions

As already described, the initial phase of an intervention for every pregnant woman

diagnosed with diabetes during pregnancy starts with nutritional and movement therapy, so

13

change in lifestyle factors of the pregnant women. Clear evidence is not yet available for

suggesting a specific type of nutritional therapy to prevent adverse events. This was

examined by a Cochrane review in 2013 (Han & Crowther, 2013). Another Cochrane review,

including not only nutritional but 15 clinical trials focusing on lifestyle intervention, showed

an effect on the outcome parameter result. The review stated that macrosomia was reduced

to a relative risk of 0.60 [95% CI 0.50; 0.71] (based on six trials including 2994 babies) when

compared to women not receiving lifestyle interventions. However, outcome parameters

including preeclampsia, caesarean section and neonatal hypoglycaemia were not reported as

significant (Brown et al., 2017). In the updated S3-guidelines of 2018, lifestyle intervention

is still regarded as the primary therapy strategy (DGG & DGGG, 2018, p. 42).

Insulin

The next treatment escalation, when glucose target levels are not met by lifestyle

interventions, is the treatment with an insulin therapy. The algorithm, which is used to define

when an escalation is necessary, was formed on the interventional MFMU trial. Hereby, all

women within the interventional group received insulin therapy, if more than 50% of blood

glucose measurements were above 95mg/dl fasting or above 120mg/dl 2h postprandial. The

measurements were performed by daily self-monitoring. Throughout the trial, 37 out of 485

women of the interventional group received insulin treatment (7.4% of the population).

Different outcome parameters were significantly reduced by the intervention, including birth

weight above 4000g (RR 0.41 95% CI [0.26 – 0.66]), caesarean delivery (0.79 [95% CI 0.64

– 0.99]), shoulder dystocia (RR 0.37 95% CI [0.14 – 0.97]), and preeclampsia (0.63 [95%

CI 0.42 – 0.96] (Landon et al., 2009). The treatment algorithm is still included in the updated

guidelines of 2018 (DGG & DGGG, 2018, p. 46). Furthermore, the guideline of 2011

specifies that insulin treatment should only be initiated by a diabetologist (Kleinwechter et

al., 2011, p. 46).

Oral hypoglycaemic agents

The use of other agents which normalize blood glucose levels, such as glibenclamide or

metformin, are not indicated for pregnancy. Glibenclamide is specifically forbidden during

pregnancy, whereas metformin is not approved and would need to be prescribed as “Off-

Label” (Kleinwechter et al., 2011, p. 50). Other guidelines, such as the guideline of the

14

National Institute for Health and Care Excellence (NICE), describe metformin as an

alternative to insulin for women with GDM (NICE, 2015, p. 62).

A Cochrane review, which compared insulin to oral hypoglycaemic agents, e.g. metformin,

concluded that there are no differences in the short-term effect of either treatment. However,

data for long-term outcomes are missing (Brown, Grzeskowiak, Williamson, Downie, &

Crowther, 2016).

3.4.4. Implications for the delivery and follow-up

The S3-Guidelines describes a birth weight of above 4500g as a weight to recommend a

section for women with a GDM. However, the guidelines recognised the uncertainty of this

value and, therefore, doctors should clearly discuss the risk and consequences of a section

with their patients (Kleinwechter et al., 2011, p. 58).

As mentioned in the epidemiology section, women with a GDM are at an increased risk for

diabetes mellitus. Therefore, the S3-Guideline recommends a follow-up 75g OGTT fasting

and after 2h in 6-12 weeks after pregnancy. To continuously monitor the manifestation of a

diabetes mellitus, an additional follow up is described. Women with an already impaired

glucose tolerance should be tested once a year with a 75g OGTT. Women with no

impairment after pregnancy should be offered the test every 2-3 years (Kleinwechter et al.,

2011, p. 61).

3.5. Health care service research reports in Germany

This analysis based on health claims data is trying to evaluate the implementation of the

guidelines of 2011. Other work in the field of health care service research have already

investigated different aspects of the health service provided for women with GDM during

pregnancy. These different reports will be represented briefly in order to describe the current

knowledge available and to identify the existing research gap.

The introduction of the two-phase general screening for pregnant women was analysed by a

qualitative study from a gynaecologist’s point of view. Hereby, 17 gynecologists were asked

to give their opinion on the new directive. Most of the gynecologists that were questioned

made use of both tests. Only a few gynecologists asked patients to self-pay the 75g OGTT

15

to avoid the application of the 50g GCT, which was seen by the doctors as not informative

enough for diagnosing the GDM. Furthermore, some gynecologists avoided to do the

confirmatory test themselves and referred women directly to a diabetologist. Only a few

women did not participate in the screening, as they actively refused the test or missed

appointments. Almost all gynecologists cooperated with diabetologists or endocrinologists.

However, gynecologists stated that they had the impression that sometimes diabetologists

were inexperienced in counselling pregnant women (Diehl et al., 2016).

Another qualitative report by the same research team described the view of 20 pregnant

women on the screening of the GDM. The report stated that, on the one hand, most women

had a positive attitude towards the screening. On the other hand, they criticized the two-stage

approach, when tested positively at the pre-test because they had to wait in fear until they

were tested with the confirmatory test. In nearly all cases screening was actively proposed

by the gynaecologists within the specified time frame. Still, there were cases where the

gynaecologist advised the patients not to do the screening at all or to avoid the pre-test (Görig

et al., 2015).

Apart from qualitative studies, also quantitative studies researched the implementation of

the GDM guideline of 2011. A group of researchers from the National Association of

Statutory Health Insurance Physicians (Kassenärztliche Bundesvereinigung; KBV) used

outpatient data from the year 2015 to estimate the prevalence of GDM and furthermore the

implementation of the screening algorithm. The screening was widely applied in more than

80% of all women. Also, when women received a test, in 95% of the cases they received at

least the GCT test. The minority of women, around 17% were tested with the confirmatory

test. The research group of the KBV estimated an administrative prevalence of the GDM of

13.2% in the year of 2015 (Melchior et al., 2017).

The perspective of the diabetologists on the screening and treatment of GDM is presented

by the registry “GestDiab”. Diabetologists from all over Germany are included and report

their experience of treating pregnant women with diabetes since 2008. Main result of the

current report of 2013/2014 was the extremely high rate of insulin initiation during

pregnancy. Around 40% of the population with GDM received insulin., much higher than

the expected 7.4% reported in the MFMU-Trial (Landon et al., 2009). There was a high

variation of insulin prescription and doses between practices. Furthermore, a low rate of

diabetes screening postpartum was stated. Only 43% of the mothers were tested postpartum

with an OGTT (Adamczewski, Weber, Faber-Heinemann, Heinemann, & Kaltheuner,

16

2016). However, it must be acknowledged, that this is only the perspective of diabetologists.

There is currently no referral rate of women with GDM, so women receiving their diagnosis

from a gynaecologist and that got successfully referred to a diabetologist. Therefore, rates

calculated by the GestDiab registry might be over- or underestimated.

Another quantitative study working with regional outpatient data to calculate the prevalence

and pregnancy complications was identified. The study identified an increased risk for all

pregnant women with GDM for hypertension, preeclampsia and caesarean section. However,

no adjustment for confounders was applied in the study. The author described that the

missing perspective on the health of the child was a limitation of the study (Tamayo et al.,

2016).

The missing perspective of the neonate was included in the work of a French team, analysing

adverse events of mother and child with data from the French hospital discharge database.

Outcomes already analysed in the HAPO study, such as caesarean section, macrosomia and

pre-eclampsia were increased in women diagnosed with GDM in comparison to women

without. Even after adjusting for common confounders, such as age and birthweight.

However, it needs to be considered, that the French recommendations in screening for the

GDM are different in comparison to Germany. The 75g OGTT is only offered in France for

women with an age above 35 years, a BMI > 24 kg/m² or women that had previous pregnancy

with GDM (Billionnet et al., 2017). Rates of adverse events for Germany might differ, due

to a population screening concept, whereas in France a population at risk screening was

implemented.

The follow-up rate with a 75 OGTT after being diagnosed with GDM was not evaluated in

the German health care system. Research from other countries, such as the US, report low

postpartum follow up rates. In a regional medical centre in the US, a low follow-up rate of

23.4% at 6-months after pregnancy was observed (McCloskey, Bernstein, Winter, Iverson,

& Lee-Parritz, 2014). In a randomized controlled trial reminder were tested to increase

follow-up rates. In the group without any reminder an one-year follow-up rate of 14.1% was

calculated. When postal reminders were sent to doctors and patients the rate increased to

60.5% (Clark, Graham, Karovitch, & Keely, 2009). No general reminding system exists

currently in Germany. Therefore, low rates for Germany can be expected, however no

research investigating a follow-up rate is available for the German health care system yet.

17

All presented health care service research reports present valuable insights into the provision

of care for pregnant women. However, certain aspect to fully understand the care situation

in Germany are missing. The cooperation between gynaecologists and diabetologists were

mentioned in the qualitative report of Diehl et al.. Still, no quantitative study focused on the

cooperation between the two specialty groups is available yet. The missing aspect of

collaboration between the two specialty groups makes the analysis of the GestDiab registry

difficult to interpret, as only women diagnosed by diabetologist are described. Furthermore,

there is no current evaluation of adverse pregnancy events in women with a GDM available

in Germany, while there is a report in France. The analysis of Tamayo et al. was able to

analyse adverse event of mothers based on outpatient data (Tamayo et al., 2016). However,

many adverse events, e.g. neonatal hypoglycaemia, shoulder dystocia and

hyperbilirubinemia, are apparent in the newborn. Further, no health care service research in

Germany focused on the follow-up patients, which is an important part in provision of care

for pregnant women diagnosed with an GDM.

Consequently, this analysis will include the named missing aspects of service research, the

cooperation between the two speciality groups, the evaluation of adverse events in neonates

and the follow-up postpartum.

18

4. Methods

In this chapter the database, the analysis plan and further specification of the study will be

described. In the first part a general overview of research with administrative database is

given. In a second part the analysis plan will be described, including chosen outcome

parameters and inclusion as well as exclusion criteria.

4.1. Health care service research with SHI - Health claims data

This epidemiological study utilized a health claims database, also referred as health care

utilization database or administrative database, to evaluate the research questions. In general,

these databases have been designed primarily for machine-readable invoices between the

care providers, e.g. doctors and hospitals, and the insurance companies. However, in recent

years databases have been reused for scientific investigations (Schneeweiss & Avorn, 2005).

In Germany, a law that insists on machine-readable invoices between hospitals and SHIs,

stated by the paragraph §301 of the Social insurance code V (Sozialgesetzbuch V), went into

effect in 2004. Since then, all hospital administrations had to be claimed within the German

diagnosis related groups (G-DRG) system. Primary and secondary diagnoses defined by the

International Classification of Diseases 10th revision in a modified German version (ICD-

10-GM), the Operationen und Prozedurenschlüssel (OPS) a German specific code for

medical procedures, number of hospital days, age and sex of the patient are used to generate

DRGs and hence invoices for insurance companies. Comparable laws and the defined

classifications systems are available for other sectors of the German health care system. The

application of standardized classification systems is hereby of importance, while only then

data may be aggregated to analyse epidemiological key measures, e.g. prevalence and

incidence, or even health care utilization, e.g. costs or numbers of doctor appointments

(Ohlmeier et al., 2014).

Apart from the hospital sector, the outpatient pharmacy and the outpatient care sector will

be utilized in this study. Main aspects of this sector will be shortly presented. Comparable

to the inpatient sector, the outpatient doctor is obligated to use the ICD-10-GM to code

diagnoses. However, the limitation is given that diagnoses are transmitted by the regional

Associations of Statutory Health Insurance Physicians (Kassenärztliche Vereinigungen; KV)

to the SHI on a quarterly basis. Services provided by outpatient doctors are accounted for by

19

the doctor’s fee scale (Einheitlicher Bewertungsmaßstab; EBM), in difference to the

diagnoses, the exact date of provision is available. Furthermore, the type of doctor specialist

is defined within the database (Ohlmeier et al., 2014; Schreyögg & Stargardt, 2012).

Outpatient pharmacies are obliged to use the anatomical therapeutic chemical (ATC)

classification system, a world health organization (WHO) standard, that has been redefined

for the German health care system, to describe the active substance of the prescription.

Additionally, the defined daily dosage (DDD) is part of the ATC classification system, to

define the package size. Apart from classification, the date of prescription is necessary to

invoice pharmaceutical prescriptions as an outpatient pharmacy. Finally, all utilized services

of insured person can be linked via an unique identification number (Ohlmeier et al., 2014).

The database this analysis was conducted on is the pseudonymized SHI database of the

Techniker Krankenkasse (TK) with approximately 10 million insured persons in 2018

(approximately 10% of the German population). Many different studies with different study

designs have already conducted research on the TK database (Lange et al., 2015; Schneider,

Linder, & Verheyen, 2016).

4.2. Identification of study groups

To be able to answer the research questions a longitudinal study design with three different

time periods and groups was developed. In general, all pregnant women continuously

insured for the time of observation, between 14 and 50 years old with a diagnosis code of

ICD-10-GM or an OPS-2016 code indicating a single or multiple delivery in 2016 were

included into the study. Hereby, a list of codes prior developed and published by a team of

health claims data scientists was used (Mikolajczyk, Kraut, & Garbe, 2013) (See codes in

Appendix) . Excluded were women with a pre-term birth before the 20 weeks of pregnancy

via the ICD-10-GM codes of O09.0! – O09.2! women, as the test for GDM is first applied

at 24th week of pregnancy. No further women were excluded, for the first part of the analysis

(t0), the application of the diagnostic tests for GDM.

For the main part, the analysis of adverse events of mother and child during the

hospitalization of delivery (t1) and the follow up (t2), only women that were either having a

reimbursed test (50g GCT/75g OGTT) for the GDM or women having a diagnosis of GDM

and prescribed self-monitor stripes. This was considered to ensure a valid diagnosis.

20

Three distinct groups of pregnant women were identified. At first, women without a

diagnosis coded inpatient or outpatient of a GDM (ICD-10: O24.4) during their time of

pregnancy. This group represents the general population of women that took part in the

screening. The second group was defined as group with an out- or inpatient diagnosis of

GDM (ICD-10: O24.4), but without a pharmacological insulin treatment. The third group

was determined by women with a GDM diagnosis and a pharmacological treatment with

insulin (ATC: A10A). In both GDM groups, women were excluded having a coded diagnosis

of diabetes mellitus type 1 or type 2 (ICD-10-GM: E10 / E11) or women who received

insulin (ATC: A10A) within one year prior to the estimated start of the pregnancy.

4.3. Periods of the analysis and definition of outcome parameters

The analysis is differentiated into three periods with the goal to evaluate the distinct aspects

of the research question (see Figure 1). In period t0: time of pregnancy, the test application

and the prescription of insulin by doctors was investigated. In t1: Hospitalization of delivery,

the incidence of adverse events of mother and child of the three groups were evaluated. The

last period was designed to investigate a follow-up of women with a diagnosis of GDM.

Primary focus for this period was to calculate the utilization rate for a glucose test following

a pregnancy with GDM. Secondarily, it was of interest to calculate the transition rate from a

GDM to a DMT1 or DMT2 resulting from the applied test.

21

Figure 1 Periods of the analysis

Objective: Evaluation of test and insulin application and the doctors involvementParameters:

→ Which doctors applied the tests?

→ How often are self-monitoring stripes prescribed?

→ Which doctors diagnosed GDM?

→ Which doctors prescribed insulin?

Objective: Evaluation of the follow-up period (1-year)

Parameters:

→ Is a follow-up test applied?

→ Which doctors applied the test?

→ Transistion rate GDM to Diabetes Mellitus

Objective: Evaluation of HAPO outcomes of different treatment patterns in comparison to the general population.Parameters:

→ Results of the HAPO outcomes

4.3.1. Period of pregnancy "t0"

Period t0 is defined as the time of pregnancy and operationalized by the date of delivery

minus the time of pregnancy. Date of delivery was defined by the date the OPS-2016 code,

that indicated the delivery, was coded. The same codes used as inclusion criteria were used

to indicate the date of birth. If no OPS-code was available, the date of delivery stated within

the hospital discharge information was used instead. The assumption has been made that

OPS-codes are coded more accurately, as they are relevant for the remuneration of the

hospitalization. Time of pregnancy was defined by a coded ICD-10-GM O09.3 – O09.7

during the hospitalization of delivery. These codes specify the length of pregnancy in a

period of days, e.g. ICD-10-GM O09.3. defines a period of 253 to 287 days of pregnancy.

To approximate the start of pregnancy, the average between the start and the end of each

interval was taken. For example, ICD-10-GM O09.5 indicates 232 to 252 days of pregnancy.

This would result in 242 days for period t0. Codes were only considered, if coded during

hospitalization. If none of these four codes were used during the hospitalization of delivery

22

a normal pregnancy with an average time of 280 days was assumed (Bergsjø, Denman,

Hoffman, & Meirik, 1990).

Main outcome of this period was the percentage of women receiving any of the screening

tests for the GDM. This was operationalized with the two specific EBM codes 01776 and

01777. EBM 01776 describes the pre-test/ 50g GCT and the EBM 01777 describes the

confirmatory test/ 75g OGTT test for the GDM. Both EBM codes were prior used in a health

services research study (Melchior et al., 2017). It was further analysed in which period of

pregnancy the tests were utilized. While the length of pregnancy is only assumed, an error

statement of 2 weeks was calculated. Additionally, it was observed how many women

received blood glucose test stripes for self-monitoring GDM. Test stripes were identified

using the ATC V04CA.

Another important measure of the first period was the diagnosis of GDM. First, the

prevalence with different validity criteria was calculated. Four level of validity were defined,

first the raw prevalence, which takes all diagnoses into account, secondly only diagnoses

with at least one test (pre- or confirmatory test) or prescribed blood glucose test were

considered. Thirdly, the highest degree of validity was assigned, if a diagnosis was coded

with a confirmatory test or the prescription of blood glucose test stripes. Finally, it was

possible to calculate a range of the GDM prevalence for Germany.

In addition to the prevalence, it was of interest which doctors coded the GDM diagnosis

(ICD-10: O24.4) at least once. Specialties of doctors were identified via the last two digits

of the lifelong doctor identifier (Lebenslange Arztnummer; LANR) given out once by the

regional KV (Kassenärztliche Bundesvereinigung, 2015). Unfortunately, diabetology is not

defined as a specialty group within lifelong doctor identifier but is described by the S3

guideline as an important group of doctors for the diagnosis and treatment of GDM

(Kleinwechter et al., 2011, p. 39). Therefore, an identification algorithm was developed for

this group of doctors.

Diabetologist are either general practitioner (GP) or internist with a certified training

organized by the DDG (DDG, 2013). The certificate is not generally needed for

reimbursement purposes, so that doctors are not easily identifiable, with two exceptions. One

is the EBM 02311 and the other the disease management program (DMP) for diabetes

mellitus type 1. The EBM 02311 specifies the counselling of patients with a diabetic foot

and can only be coded by orthopaedist or GP and internists with a certified training in

23

diabetology by the DDG (Kassenärztliche Bundesvereinigung, 2016). All GP’s, taking part

in DMP for diabetes mellitus type 1, an integrated care program for chronic disease, are

required of having a certified training in diabetology (G-BA, 2004). Therefore, doctors were

identified as diabetologist or GP/internist with special training in diabetology, if they either

coded the EBM 02311 or took part in the DMP for DMT1 within the year of 2016. All other

important doctor identifiers were grouped to represent specialist groups. The groups used

can be seen in Table 2.

Table 2 Doctor specialty groups

Group name Specialty code

Diabetology/GP with diabetes specific training

certificate

01, 02, 03: General practitioner

23: Internal medicine

25: Endocrinology

AND part of the diabetes type 1 DMP

OR coded EBM 02311

GP without additional training certificate 01, 02, 03: General practitioner

23: Internal medicine

25: Endocrinology

AND not part of the Diabetes Type 1 DMP

OR coded EBM 02311

Gynaecology 15: Gynaecology and obstetrics

16: Gynaecology endocrinology and reproductional

medicine

17: Gynaecological oncology

18: Special obstetrics and perinatal medicine

Laboratory medicine 48: Laboratory medicine

Nephrology 29: Nephrology

Angiology 24: Angiology

Others All other doctor groups not mentioned

The third outcome of the period t0 was the prescription rate of insulin for women with a

diagnosed GDM. Insulin prescription was defined by a coded ATC code A10A. Comparable

to the diagnosis of GDM it was further of interest, which specialty group prescribed insulin

during pregnancy. The same doctor specialty group scheme from Table 2 was applied.

4.3.2. Period of delivery "t1"

The period of delivery is the primary part of the analysis and contained all major outcome

parameters to answer the main research question. All pregnant women without a test of

24

screening/self-monitor stripes were excluded for this part and further parts of the study.

Furthermore, only mothers with a linkage to their newborn were included into the second

part, as outcomes were mother, and child related. The linkage of the mother to the child is

described in the section 4.4.

The baseline characteristics and outcomes chosen were part of the hyperglycaemia and

adverse pregnancy outcomes (HAPO) study (see Table 3) (Metzger et al., 2008). Maternal

baseline characteristics were: maternal age, multiple gestation, obesity defined by the world

health organization (WHO), and hypertension. Foetal baseline characteristics were age of

gestation in weeks at delivery and sex.

All clinical outcomes of the HAPO study were operationalized for German health claims

data. An overview can be seen in Table 3. The operationalization was possible in all

parameters, without the measurement of C-peptides in the cord-blood serum.

Table 3 Operationalization of HAPO outcomes

HAPO outcome variables Classification Operationalization in German health claims data

Primary outcome

Birth weight >90th percentile ICD-10-GM P08.0 Exceptionally large baby (birthweight above

4500g)

P08.1 Other heavy for gestational age infants

or birth weight above 4000g according to the discharge

information

Primary caesarean section OPS-2015 5-740.0 primary section

5-741.0 primary section supra-cervical

5-741-2 primary section corporal

5-741-4 primary section, longitudinal incision

5-742.0 primary section extraperitoneal is

5-749.10 Misga-Ladach section primary

Clinical neonatal

hypoglycaemia

ICD-10-GM P70.0 syndrome of the child of a gestational diabetic

mother

P70.1 syndrome of the child of a diabetic mother

P70.2 Neonatal diabetes mellitus

P70.3 Iatrogenic neonatal hypoglycaemia

P70.4 Other neonatal hypoglycaemia

Cord-blood serum C-peptide

above 90th percentile

Not applicable

25

Table 3 (continued)

Secondary outcome

Premature delivery (before

37 wk)

ICD-10-GM O09.0! - O09.5! Premature delivery before 37 weeks

Shoulder dystocia or birth

injury

ICD-10-GM O66.0: Shoulder dystocia;

O66.1 Obstructed labour due to locked twins

O66.2 Obstructed labour due to unusually large fetus

O66.9 Obstructed labour, unspecified

Intensive neonatal care OPS-2015 8-93 Monitoring of respiration and the cardiovascular

system

Hyperblirubinemia ICD-10-GM

ICD-10-GM

OPS-2015

P58.- Neonatal jaundice due to other excessive

haemolysis

P59.-: Neonatal jaundice from other and unspecified

causes

8-560.2 Phototherapy of the neonate (for

Hyperbilirubinemia)

Preeclampsia O11.-: Pre-eclampsia superimposed on chronic

hypertension

O14.- Pre-eclampsia

4.3.3. Period of follow-up "t2"

In the guidelines for the gestational diabetes mellitus it is recommended to test for glucose

tolerance impairment 6-12 weeks after giving birth (Kleinwechter et al., 2011, p. 61).

Therefore, this period was designed to evaluate the application of a follow-up test. As in

period t0 doctors billing the test were of interested, therefore the same group was used (see

Table 2). The EBM codes to identify the application of a glucose tolerance impairment were

32057 and 32025 (Measurement of glucose). Based on this information the rate of test

application within 6, 12, 18, 24 weeks and one year after giving birth were calculated. In

addition, the 1-year transmission rate from GDM to DMT1 or DMT2 will be assessed in the

one-year individual follow-up after delivery, to ensure a validity of the diagnosis.

4.4. Linkage of mother and child

The health of the newborn child must be considered, when trying to evaluate the

diagnostic/treatment of GDM in Germany, while many outcome parameters of the HAPO

outcomes are child specific (Metzger et al., 2008). However, as the database is

26

pseudonymised and family relations are not necessary for billing purposes, no family linkage

is generally available. A method to link mothers to their newborn has been established by

other researchers working with German statutory health insurance (SHI) data. The

researchers identified three possible ways on how the newborn is insured in the SHI (see

Figure 2).

Firstly, the newborn is co-insured on the mother’s insurance (direct linkage), here the mother

needs to be insured as a member. Mother and child can be directly linked by using the

pseudonymized member identification number. Secondly, the newborn and the mother are

co-insured on the father’s insurance (indirect linkage). The father is the main member of the

insurance and can be identified using the member identification number stated within the

insurance record of the mother. In a second step the mother can be indirectly linked to her

child via the father’s member identification number. Thirdly, the child was co-insured on the

father insurance, but the mother was not. Then a linkage of mother and child is not possible

(Garbe, Suling, Kloss, Lindemann, & Schmid, 2011). The same technique has been used

within this study to identify mother-baby pairs.

Figure 2 Linkage of mothers to their baby in SHI databases

27

4.5. Validation of the "obstetric comorbidity index" in a German setting

In the analysis it was assumed that generally differences in the prevalence co-morbidities

between the groups will occur, as the group of mother with GDM have a higher mean age

(Melchior et al., 2017). A difference in the prevalence of co-morbidities would have

confounded the outcome parameter of t1. Therefore, the idea was to use a comorbidity index

in a regression analysis to adjust for the difference in the comorbidity burden. Comorbidity

scores are generally used to adjust for confounding by comorbidities in administrative

database studies. Scores are usually based on multiple comorbidities. Each comorbidity

receives a value based on the association to a severe outcome parameter, for example death.

Association might be measured by odds ratio, relative risk or an expert panel. The higher the

association with the outcome the higher is the assigned value. An individual score is then

calculated by summarizing all values into a single score. The higher the final score the more

likely is the incidence of the final outcome parameters (Schneeweiss & Maclure, 2000). The

two most known comorbidity scores are the Charlson Comorbidity Index and the Elixhauser

Score (Elixhauser, Steiner, Harris, & Coffey, 1998; Van Walraven, Austin, Jennings, Quan,

& Forster, 2009). While these indices are based on the prediction of mortality and therefore

intended for an older more morbid population, they are not applicable in an obstetric

population (Bateman et al., 2013). A systematic review performed by Aoyama et al. (2017)

recognized the "Comorbidity Index for Use in Obstetric Patients" invented by Bateman et

al. as the only instrument to measure the comorbidity burden in a pregnant population

(Aoyama, D’Souza, Inada, Lapinsky, & Fowler, 2017; Bateman et al., 2013). Therefore, the

index was chosen for the adjustment of comorbidities. However, the index has never been

used for a German health claims database and therefore needs to be validated for those.

The final index consists out of 20 conditions (see Appendix) and age and is based on the

prediction of maternal end-organ damage (Bateman et al., 2013). The initial index was

designed for the ICD-9th revision. To make the index applicable for other non-US

administrative health databases, a team of Canadian researcher updated the index to the ICD-

10th revision and validated it for the Canadian health care system (Metcalfe et al., 2015).

In order to validate a comorbidity index, usually the area under the curve (AUC) of the

receiver operator characteristic curve (ROC) is calculated (Schneeweiss & Maclure, 2000).

The ROC curve can be seen in Figure 3. On the y-axis the sensitivity (i.e., true positive rate)

is plotted and on the x-axis the 1-specificity (ie, true negative rate). The ROC, curve B in

28

Figure 3, shows different states for the sensitivity and specificity for an infinite number of

cut-off values (Zou, O’Malley, & Mauri, 2007). AUC is the discrimination statistic for

dichotomous variables and is available for the logistic regression.

It compares the predicted outcome, e.g. death or end-organ damage, to the actual occurrence

of the outcome. Hereby, a final value between 0 and 1 is calculated. A value of 0,5 meaning

a prediction by chance and 1 would mean a perfect prediction of the outcome (Schneeweiss

& Avorn, 2005). The updated Elixhauser Score reached an AUC of 0,763 [95% CI; 0,76 -

0,77] in an Canadian population with death as outcome (Van Walraven et al., 2009). Also

the ICD-10th revision update by Metcalfe reached an comparable, but lower AUC of 0,70

[95% CI; 0,60 - 0,80], when using the obstetric comorbidity index with end-organ damage

as an outcome (Metcalfe et al., 2015).

To assess the validity of the obstetric index (OCI) for German health claims data the

discriminatory power was compared to the Elixhauser Score in the population of the t0

period. As performed by Bateman et al. (2013) the comorbidities of the OCI were extracted

during the hospitalization of delivery and for the Elixhauser all healthcare contacts one-year

prior to the delivery were considered. The outcome of end-organ damage was assessed

during the hospitalization of delivery + 30 days after hospital discharge (Bateman et al.,

2013).

Figure 3 Receiver operator characteristic curve by (Zou et al., 2007)

29

Due to specific reasons that were visible within the validation process, the obstetric

comorbidity index was excluded from the main part of the analysis. Interested readers can

get insights to the results, the discussion and reasons why the index was excluded from the

analysis within the excurse section (chapter 8).

4.6. Statistical analysis

All parameters calculated in the periods "t0" and "t2" were descriptive. Mean and standard

deviation were calculated for continuous variables and percentages for categorical variables.

Phi-coefficient was used as a measure of strength association, in cross-tabulation established

within the two periods, t0 and t2. Cramer’s V was used to identify an association between a

dichotomous and ordinal scaled variable. Chi-square test was used to test for general

association of two categorial variables. In the main part of the analysis t1, odds ratio (OR)

and the 95% Confidence Interval (95% CI) were calculated for categorical variables. The

group not diagnosed with GDM served as reference group. In addition to the unadjusted

ORs, multiple regression models will be fitted for each outcome parameter. For maternal

outcomes a multiple logistic regression model including stratified age groups, three level of

obesity defined by the WHO, and multiple gestation was fitted. For neonatal outcomes the

same potentials confounders and the sex of the neonate was included into a logistic

regression model. The analysis was performed using SAS Enterprise Guide 9.3.

30

5. Results

In total 109,568 women who gave birth in 2016 were identified within the insurance

database. From those 21,825 were excluded from the analysis, because there were not

continuously insured, without continuous residence within Germany for the whole

observational period or had a pre-term birth before the 20nd week of pregnancy. Other

reasons can be identified in Figure 4. From the initial sample, 88,632 were eligible for the

first part of the study and the validation of the obstetric comorbidity index (see chapter 8:

Excurse). For the next period, the incidence of adverse pregnancy outcomes, the sample was

further reduced. Women without at least one screening test for GDM or at least one

prescription of blood glucose stripes were excluded for the whole period of t1 and t2, which

resulted in an exclusion of 14,199 women or 17.6% of the initial study population. Women

with a test were separated for the three study groups, general population, GDM without

insulin treatment and GDM with insulin treatment. Of the 73,957 women eligible for the

second part, 63,031 were assigned to the general population, while having no present

diagnosis of GDM coded in the outpatient sector. From the 10,926 women with a diagnosis

of GDM, 476 were excluded, due to a diagnosis of diabetes type 1/2 or a prescription of

insulin within one year prior to pregnancy. Women with GDM diagnosis were then assigned

to no insulin treatment (n=9,157) or to insulin treatment (n=1,769) in addition to their

diagnosis. In the next step, the linkage of mothers to their newborn child, 58,297 mother-

baby pairs were identified, representing 79% of the population prior to linkage. The final

group of women without GDM consisted out of 49,645 mothers. The GDM group without

insulin treatment consisted out of 7,245 mothers and the last group out of 1,407 women.

31

Figure 4 Flowchart Study population: data year 2016

Inclusion via ICD-10-GM

codes + birth as reason

for admission (n=109,397)

Inclusion via OPS-2016

codes + birth as reason

for admission (n=109,156)

Eligible pregnant women

(n=88,632)

Combination (n=109,568)

Mother-child linkage

No GDM diagnosis (ICD O24.4)

(n=63,031)

GDM diagnosis with confirmatory

test (ICD O24.4 + EBM 01777/

01776) (n= 11,402)

No Diabetes diagnosis before

pregnancy

(n=10,926)

No insulin treatment

(n=9,157)

Insulin treatment with

onset during pregnancy

(n=1,769)

Pregnant women without the

Diagnosis of GDM (n=49,645)

Pregnant women with the

Diagnosis of GDM and without

insulin (n=7,245)

Pregnant women with the

Diagnosis of GDM and with insulin

treatment (n=1,407)

Exclusions (n=21,285)

- Women not between 14 - 50 years (n=2)

- No remunerated DRG (n=19)

- Pregnancy less than 22-Weeks (n=534)

- Two labour induction within the same year (n=13)

- Uncontinously insured (n=20,140)

- Uncontinous residence in Germany (n=228)

Exclusion of women with Diabetes

Type 1/2 or insulin treatment

before pregnancy: (n=476)

t0 + validation of the obstetric

comorbidity index

Women eligible for t1

(n=74,433)

Exclusion of women without

screening for GDM or Glucose

self-monitoring (n=14,675)

32

5.1. Period of pregnancy “t0”

In this period, the focus was on the application of the screening tests, the prescription of

insulin, the diagnosis of GDM and the doctor’s involvement in those procedures. For the

period of pregnancy 88,632 women were eligible.

5.1.1. Test utilization

In overall, 82.33% (n=77,968) of the population received either the pre-test (50g GCT) or

the confirmation test (75g OGTT). Pre-test was applied in 77.01% of all pregnant women.

In 18.14% of all pregnant women the confirmatory test was used. Only a minority of the

population tested with the OGTT, received the confirmatory test without a pre-test (5.32%

of the population) (see Figure 5).

Table 4 presents the week of pregnancy in which the pre-test was applied. As the time of

pregnancy is based on assumptions, a two-week window around the estimated mean was

used. The data revealed that the middle 80%, ie the time between the 10th and 90th percentile,

of all pregnant women received the 50g GCT (pre-test) test in a short time interval of four

weeks. The estimated mean of week pregnancy, when the 50g GCT test was utilized, can be

described as the 25.45 (SD 2.13) week of pregnancy. Taking the uncertainty of the ICD-10

codes, the mean lies between the 23rd (see lower barrier) and the 27th week (see upper barrier)

of pregnancy.

15,664; 17.67%

56,891; 64.19%

11,365 ; 12.82%

4,712; 5.32%Screening test in pregnant women

No test

50g GCT without 75g OGTT

50g GCT with 75g OGTT

75g OGTT without 50g GCT

Figure 5 Screening test in pregnant women

33

Table 4 Time of test execution, 50g GCT

Variable (n=68,256) Mean (SD) 10th Percentil 90th Percentil

50g GCT, week of pregnancy

(Lower barrier)

23.45 (2.13) 21.28 25.85

50g GCT, week of pregnancy 25.45 (2.13) 23.28 27.85

50g GCT, week of pregnancy

(Upper barrier)

27.45 (2.13) 25.28 29.85

In Table 5, the same numbers for the 75g OGTT (confirmation test) are presented. For the

middle 80% of the pregnant women the test was utilized between the 23rd and 31st week of

pregnancy, a time interval of 8 weeks. The time interval where the test was utilized by

women was therefore around two times larger in comparison to the execution of the pre-test.

The mean week of pregnancy, when test was received, was 26.6 (SD 4.4) weeks. However

due to uncertainty of the assumptions the mean lies between the 24th and 28th weeks of

pregnancy.

Table 5 Time of test execution, 75g OGTT

Variable (n=16,077) Mean (SD) 10th Percentil 90th Percentil

75g OGTT, week of pregnancy

(LC)

24.65 (4.40) 20.85 29.57

75g OGTT, week of pregnancy 26.65 (4.40) 22.85 31.57

75g OGTT, week of pregnancy

(UC)

28.65 (4.40) 24.85 33.57

The distribution of the two tests over time are displayed in a histogram (Figure 6). On the x-

axis the weeks of pregnancy and on the y-axis the proportion of the population are shown.

The utilization of the 50g GCT is presented in blue and the 75g OGTT in green bars.

Distributions are displayed for the general assumption without the two weeks of uncertainty.

The graph explains the descriptive results of the tables above (see Table 4 and 5). The

application of the 50g GCT test follows a narrow normal distribution with a sharp increase

34

between the 20th and the 30th week of pregnancy. In comparison to the distribution of the

pre-test, the distribution of the confirmatory test was wider and was slightly skewed left. The

distribution peaked about two weeks shifted to the right in comparison to the 50g GCT.

Furthermore, the confirmatory test was applied all over pregnancy, starting with a small

proportion in the early days of pregnancy, before the 20th week of pregnancy, ending with a

population, which received the test from the 30th week onwards.

Figure 6 Week of pregnancy, when tests were applied

week of gestation

35

Table 6 Utilization of 50g GCT by doctor specialist groups

Field of

specialization

Frequency of 50g GCT Percentage Cumulative

frequency

Cumulative

percentage

Gynaecology 68086 99.75 68086 99.75

Diabetology/GP

with additional

training

110 0.16 68196 99.92

GP without

additional

training

55 0.08 68251 100.00

Nephrology 2 0.00 68253 100.00

Others 1 0.00 68254 100.00

In Table 6 and 7 the percentage of doctors prescribing the 50g GCT and the 75g OGTT is

displayed, respectively. From the 68,254 women who utilized the 50 GCT test, 68,086

(99.75%) received the test in a gynaecologist office. The percentage shifts in the provision

of the 75g OGTT. Gynaecologist were still the specialty group, who executed the test most

often, with a rate of 75% or 12,081 from 16,077 receiving the test in total. Second highest

specialty group was the diabetologist with a rate of 22.49%. The two specialist groups

conducted more than 97% of all tests.

Table 7 Utilization of 75g OGTT by doctor specialist groups

Field of

specialization

Frequency of 75 OGTT Percentage Cumulative

frequency

Cumulative

percentage

Gynaecology 12081 75.14 12081 75.14

Diabetology/GP

with additional

training

3616 22.49 15697 97.64

GP without

additional

training

227 1.41 15924 99.05

Angiology 64 0.40 15988 99.45

Nephrology 50 0.31 16038 99.76

Others 39 0.24 16077 100.00

5.1.2. Blood glucose test stripes

In the first part of analysis, it was not just of interest, how many took part in the screening,

but also how many received a prescription for blood glucose test stripes. Therefore, a cross-

tabulation of the two variables, prescription of blood glucose test stripes and application of

36

at least one screening test was plotted (see Table 8). In general, 7,204 women (8.13% of the

total population) received a prescription for blood glucose stripes. In 20% of women test

stripes were prescribed without an applied screening test. It is apparent from this table that,

either doctors use the blood glucose stripes as an additional screening method or that women

already established diabetes mellitus type 1 or 2 prior to pregnancy and are in general need

of test stripes. The association between an applied screening test and blood glucose self-

monitoring was significant [χ² (1) = 38.64; p<0.001]. However, strength of association was

low (ϕ = 0.02).

Table 8 Cross-tabulation: Blood glucose self-monitoring * screening

Screening (either 50g or 75g Test)

Sel

f-m

on

ito

rin

g s

trip

es d

uri

ng

pre

gn

an

cy

NO YES Total

NO

Frequency 14,199 67,229 81,428

Total, % 16.02% 75.85%

Row, % 17.44% 82.56%

Column, % 90.65% 92.13%

YES

Frequency 1,465 5,739 7,204

Total, % 1.65% 6.48%

Row, % 20.34% 79.66%

Column, % 9.35% 7.87%

Total 15,664 72,968 88,632

χ (1) = 38.64; p <0.001 ϕ = 0.02

The blood glucose test stripes were mostly prescribed by a general practitioner without any

additional training in diabetology (n= 4,946; 69.94%) (see Table 9). The frequencies of

doctor’s specialty groups prescribing blood glucose stripes were different in comparison to

the specialty conducting the screening. In the process of screening, the gynaecologist and

the diabetologist have been the main specialist group in conducting these tests.

37

Table 9 Doctor specialty group prescribing blood glucose stripes

Field of

specialization

Frequency of insulin

prescription

Percentage Cumulative

frequency

GP without

additional

training

4,946 69.94% 4,946

Diabetology/GP

with additional

training

1,317 18.62% 6.263

Gynaecology 483 6.83% 6,746

Others 135 1.91% 6,881

Nephrology 102 1.44% 6,983

Angiology 89 1.26% 7,072

5.1.3. Diagnosis of the GDM

Of the 88,632 pregnant women, 11,755 had a coded diagnosis of gestational diabetes

mellitus in the outpatient sector. However, of those 476 already were already diagnosed with

diabetes type 2, type 1 or received an insulin prescription within the year prior to pregnancy.

After exclusion, 11,279 women still had valid diagnosis of GDM. The administrative

prevalence of the GDM in Germany in 2016, when considering all diagnoses, was 12.73%

(n=11,279) in all pregnant women. Of those 10,926 had a diagnosis with at least one

screening test for GDM or a prescription for blood glucose stripes, resulting in an

administrative prevalence of 12.33%, when considering these methods for verification. The

highest level of validity of diagnosis was defined, when the confirmatory screening test was

performed, or blood glucose stripes were prescribed. The prevalence decreased to 9.87 %

(n=8,748). The final administrative prevalence for the gestational diabetes mellitus can

therefore be described as a range between 9.87% to 12.73%. The administrative prevalence

with different degrees of validity is presented in Figure 7.

38

Figure 7 Administrative prevalence with different degrees of validity

Table 10 provides an overview of which doctors coded a diagnosis of GDM for the 11,755

women with a diagnosis and at least one test for GDM. The most important specialist in

diagnosing the GDM was the gynaecologist, which diagnosed 84.96% (n=9,987) of the

women with at least one coded diagnosis of GDM. Surprisingly, GP’s with a special training

in diabetology diagnosed 52.38% (n= 6,157) of pregnant women with a GDM. Normal GP’s

without any specialization in diabetology only diagnosed 9.99% (n=1,174) of the population

with at least one GDM diagnosis. Apart from these three specialist fields, no other type of

doctors had a high rate of diagnosing the disease.

0% 2% 4% 6% 8% 10% 12% 14%

Percentage of pregnant women with a diagnosis of GDM

Administrative prevalence of gestational diabetes mellitus

with 75g OGTT / blood glucose stripes with 50g GCT test / blood glucose stripes

without test/blood glucose stripes DMT1/2 or insulin prescription prior to pregnancy

Prevalence estimate, with at least one test / blood glucsose stripes:

12.33% (n=10,926)

Prevalence, regardless of screening tests: 12.73% (n= 11,279)

Prevalence estimate, with 75g OGTT / blood glucose stripes:

9.87% (n=8,748)

39

Table 10 Diagnoses coded by doctors

Field of specialization Percentage of diagnoses coded Frequency of

diagnoses coded

Gynaecology 84.96% 9,987

Diabetology/GP with diabetes

specific training certificate

52.38% 6,157

GP without additional

training certificate

9.99% 1,174

Angiology 1.35% 124

Nephrology 1.08% 107

Laboratory medicine 0.87% 135

Others 2.91% 342

TOTAL of women with a

diagnosis

11,755

In order, to clarify the importance of the gynaecologists and diabetologists in diagnosing the

GDM in the German health care setting, a cross-table for the two specialist groups is

presented in Table 11. The table is quite revealing in different ways. Firstly, nearly all

diagnoses were coded by either one of the specialist groups. Only 2.67% (n= 314) of all

women diagnosed with GDM did not had a coded diagnosis by one of the two groups.

Secondly, a low number of 40.01% of women diagnosed with GDM had a coded diagnosis

from both specialist groups and thirdly a low number of 47.09% diagnosed by the

gynaecologist received also a diagnosis from the diabetologist. This shows that less than a

half of pregnant women with GDM received care by a diabetologist, even though they had

a diagnosis coded by a gynaecologist. A χ² -square test of independence was performed to

determine whether the observed difference in diagnosing a GDM were statistically

significant. Diabetologists were significantly less likely to diagnose a GDM, then

gynaecologists [χ² (1) = 718.2; p <0.01; ϕ = 0.25].

40

Table 11 Cross-table diagnoses by specialist groups: Gynaecology * diabetology

Gynaecology

D

iab

eto

log

y/G

P w

ith

dia

bet

es s

pec

ific

tra

inin

g c

erti

fica

te

No GDM

diagnosed

GDM

diagnosed

Total

No GDM diagnosed

Frequency 314 5,284 5,598

Total, % 2.67% 44.95%

Row, % 5.61% 94.39%

Column, % 17.76% 52.91%

GDM diagnosed

Frequency 1,454 4,703 6,157

Total, % 12.37% 40.01%

Row, % 23.62% 76.38%

Column, % 82.24% 47.09%

Total 1,768 9,987 11,755

χ² (1) = 743.9; p <0.001 ϕ = 0.25

Blood glucose stripes and diagnosis

When opposing the coded diagnosis to the self-monitoring stripes in a cross-tabulation (see

Table 12) it got apparent, that all blood glucose test stripes were prescribed in women with

a diagnosis of GDM. All 7,204 women, who received a prescription for blood glucose test

stripes prescribed, also had coded diagnosis of GDM. It can be further extracted from the

table below, that 61.28% of all women with a GDM diagnosis received a prescription for

blood glucose stripes. There was a significant strong positive correlation between the

prescription of blood glucose stripes and a diagnosis [χ² (1) = 51,281.9; p <0.01; ϕ = 0.76].

Table 12 Cross tabulation: Blood glucose stripes * GDM diagnosis

Diagnosis of GDM

Blo

od

glu

cose

str

ipes

No GDM

diagnosed

GDM

diagnosed

Total

Test stripes, no

Frequency 76,877 4,551 81,428

Total, % 86.74% 5.13%

Row, % 94.41% 5.59%

Column, % 100.00% 38.72%

Test stripes, yes

Frequency 0 7,204 7,204

Total, % 0% 8.13%

Row, % 0% 100.00%

Column, % 0% 61.28%

Total 76,877 11,755 88,632

χ² (1) = 51,281.9; p <0.001 ϕ = 0.76

41

5.1.4. Insulin prescription

In the last part of this chapter, the prescription of insulin for women with GDM is described.

Of the 10,926 women with a diagnosis of GDM whom received at least one of the screening

tests or a prescription for blood glucose test was present, a frequency of 1,378 (13.87%)

women had a prescription for insulin. Doctors prescribing the initial dose are displayed in

Table 13. A large proportion (n= 1,142; 82.87%) of women received their initial outpatient

prescription by a GP with a special training certificate in diabetology. The second specialist

group prescribing insulin on a regular basis were the general practitioner with 8.85% of all

initial prescriptions. All other specialist groups can be neglected when looking at the

prescription of insulin, as more than 90% of all initial prescriptions were filled in by the first

two groups.

Table 13 Initial prescriptions of insulin by specialist groups

Field of

specialization

Frequency of insulin

prescription

Percentage Cumulative

frequency

Cumulative

percentage

Diabetology/GP

with additional

training

1466 82.87% 1,466 82.87%

GP without

additional

training

157 8.85% 1,623 91.72%

Gynaecology 50 2.83% 1,673 94.55%

Nephrology 13 0.73% 1,686 95.28%

Angiology 26 1.45% 1,711 96.73%

Others 24 1.38% 1,736 98.11%

Not defined 33 1.89% 1,769 100%

As the diabetologist was the most important doctor in prescribing insulin to women with a

diagnosis of gestational diabetes mellitus, a 2*2 table was displayed to investigate the

relationship further (see Table 14). In the row, percentages of women receiving insulin are

displayed, whereas in the column women diagnosed/not diagnosed by diabetologist are

displayed. When diagnosed by diabetologist, women receive in 26.69% of all cases an

insulin prescription. When not diagnosed by a diabetologist, this number drops to 4.78%.

This relationship was statistically significant (χ² (1) = 897.65; p <0.001 ϕ = 0.29).

42

Table 14 Cross-table Insulin * Diagnosis by a diabetologist

Prescription of insulin

D

iab

eto

log

y/G

P w

ith

dia

bet

es s

pec

ific

tra

inin

g c

erti

fica

te

No Insulin Insulin Total

No GDM coded by

specialist

Frequency 5,131 303 5,434

Total, % 46.96% 2.77%

Row, % 94.42% 5.58% Column, % 56.03% 17.13%

GDM coded by

specialist

Frequency 4,026 1,466 5,492

Total, % 36.85% 13.42% Row, % 73.31% 26.69%

Column, % 43.97% 82.87% Total 9,157 1,769 10,926

χ² (1) = 897.65; p <0.001 ϕ = 0.29

The last table of this section is a cross tabulation, investigating the association between the

two variables, prescription of insulin and prescription of blood glucose stripes. From the

chart it can be drawn, that all women who received insulin, also received blood glucose test

stripes. It was of further interest, that 56.20% of women not receiving insulin, received blood

glucose test stripes.

Table 15 Cross-tabulation: Insulin * blood glucose stripes

Prescription of insulin

Blo

od

glu

cose

str

ipes

No Insulin Insulin Total

Test stripes, no

Frequency 4,011 0 4,011

Total, % 36.71% 0%

Row, % 100% 0%

Column, % 43.80% 0%

Test stripes, yes

Frequency 5,146 1,769 6,915

Total, % 47.10% 16.19%

Row, % 74.42% 25.58%

Column, % 56.20% 100.00%

Total 9,157 1,769 10,926

χ² (1) = 1224.32 p <0.001 ϕ = 0.33

43

5.2. Period of delivery “t1”

This chapter represents the main part of the analysis, focusing on adverse pregnancy

outcomes during the hospitalization of delivery. For this part, only mothers screened for

GDM and that were further linkable to their newborn were included. The deviation of the

study sample can be seen in Figure 2. For Group A, women with no GDM, 49,645, for Group

B, women with GDM, but without the prescription of insulin, 7,245 and Group C, women

with GDM and prescription of insulin, 1,407 women were eligible. Table 16 provides an

overview of the baseline characteristics at the time of the hospitalization of the three study

groups. At first glance, there are many aspects interesting at this table. Even though, the

mean age is not largely different between the groups and in a range of 1.7 years (Group A:

32.3 – Group C 34.1 years), proportions of women above 35 years of age increased with the

diagnosis of GDM and is even more increased in women treated with insulin. In Group A, a

proportion of 25.9% and in Group C 38.2% had an age above 35 years. Increased proportions

were observed in all variables in Group C. For example, the percentage of women with

chronic hypertension was three times higher in Group C in comparison with Group A. In

summary, women with a GDM tend to have a higher age, were tested more frequently with

a confirmation test and had higher rates of comorbidities. In women treated additionally with

insulin, observed variables are even more increased. In obesity and hypertension, the

proportion even doubled in women with the treatment of insulin treatment in comparison to

women without.

44

Table 16 Baseline characteristics

Group A

(No GDM)

Group B

(GDM)

Group C

(GDM with Insulin)

Number 49,645 7,245 1,407

Age, mean (SD) 32.28 (4.47) 33.07 (4.56) 34.06 (4.23)

Age betw.35-39, n (%) 12,898 (25.98%) 2,236 (30.86%) 538 (38.24%)

Age above 40, n (%) 2,520 (5.08%) 541 (7.47%) 139 (9.88%)

Screening/Blood glucose

stripes

75g OGTT, n (%) 7,760 (15.69%) 3,919 (54.09%) 892 (63.40%)

Blood glucose stripes 0 (0.00%) 3,959 (54.64%) 1,407 (100%)

Multiple gestation, n (%) 856 (1.72%) 160 (2.21%) 21 (1.49%)

Hypertension

Chronic, n (%) 2,727 (5.49%) 604 (8.34%) 203 (14.43%)

Gestational, n (%) 1,492 (3.01%) 343 (4.73%) 106 (7.53%)

Preeclampsia, n (%) 2,852 (5.74%) 586 (8.09%) 139 (9.88%)

Obesity

TOTAL, N (%) 2208 (4.45%) 740 (10.21%) 331 (23.50)

WHO Grade 1, n (%) 1009 (2.03%) 319 (4.40%) 101 (7.19%)

WHO Grade 2, n (%) 731 (1.47%) 238 (3.28%) 127 (9.02%)

WHO Grade 3, n (%) 468 (0.94%) 183 (2.53%) 103 (7.29%)

As not only the health of the mother is reflected by this study, but also the health of the

newborn, general characteristics of the neonates are summarized in Table 17. For this study,

50,492 for Group A, 7,405 for Group B and 1,426 babies for Group C were linked to their

mothers. In all three groups, the proportion of female newborns was slightly less than 50%.

When looking at the week of gestation the baby was born, an interesting observation was

made. The number of babies born between the 37 – 41 weeks increased from Group A to

Group C, whereas the proportion of babies born after the 41 week of gestation decreased

form Group A to C. Group A, B and C had percentages of babies born after the 41 weeks of

gestation of 11.30%, 9.08% and 1.26%, respectively. Preterm birth, between weeks 34-36,

was slightly increased in Group C. However, the proportion of earlier preterm birth was

lower in comparison to the other two groups.

45

Table 17 General characteristics of the newborn

Group A

(No GDM)

Group B

(GDM)

Group C

(GDM with Insulin)

Number 50,492 7,405 1,426

Female, n (%) 24,825 (49.17%) 3,569 (48.20%) 665 (46.63%)

Week of gestation at birth

20 – 25 weeks, n (%) 19 (0.04%) 1 (0.01%) 1 (0.07%)

26 – 33 weeks, n (%) 782 (1.55%) 124 (1.67%) 16 (1.12%)

34 - 36 weeks, n (%) 1,866 (3.70%) 352 (4.75%) 70 (4.91%)

37 - 41 weeks, n (%) 42,121 (83.42%) 6,272 (84.70%) 1,324 (92.85%)

More than 41 weeks, n (%) 5,704 (11.30%) 656 (8.86%) 15 (1.05%)

The raw frequency of the primary outcomes (birth weight <90th percentile, clinical neonatal

hypoglycaemia, primary caesarean section) are presented in Figure 8. In all three graphs, the

three groups are displayed on the x-axis, and the proportion of each group affected on the

y-axis. In general, all primary outcomes increased over the three obtained populations.

However, they increased with different rates.

Figure 8 Frequency of primary outcomes across study groups

Clinical neonatal hypoglycaemia increased nearly exponential between the groups. Group A

had a frequency of 1.96%, Group B of 4.82% and Group C of 10.24%. Primary section,

46

however, increased on a low rate between Group A and B and had a strong increase from

Group B to C. Rates were 15.90% for Group A, 19.16% for Group B, and 27.43% for Group

C. The same observation can be drawn for the outcome of birth weight. The frequency

increased from 1.46% to 1.76% and then to 3.69% in Group C. The frequency of all

outcomes, secondary and primary, are displayed in Table 18. In all outcomes, secondary and

primary, with exceptions of bilirubinaemia and premature delivery, Group C had the highest

frequency (see Table 18).

Table 18 Frequency of the HAPO Outcomes by groups

HAPO Outcomes Group A

(No GDM)

Group B

(GDM)

Group C

(GDM with Insulin)

Neonatal outcomes n= 50,492 n= 7,275 n= 1,426

Birth weight >90th percentile 737 (1.46%) 130 (1.76%) 55 (3.69%)

Clinical neonatal hypoglycaemia 990 (1.96%) 357 (4.82%) 146 (10.24%)

Premature delivery 2667 (5.28%) 477 (6.44%) 87 (6.10%)

Shoulder dystocia or birth injury 277 (0.55%) 33 (0.53%) 9 (0.63%)

Intensive neonatal care 5,101 (10.10%) 894 (12.07%) 187 (13.11%)

Neonatal Bilirubinaemia 2,444 (4.84%) 449 (6.06%) 63 (4.42%)

Maternal outcomes n=49,645 n= 7,245 n= 1,407

Primary section 7,893 (15.90%) 1,388 (19.16%) 386 (27.43%)

Preeclampsia 2,852 (5.74%) 586 (8.09%) 139 (9.88%)

Logistic regressions for the primary outcome parameters

To describe the likelihood of occurrence of the outcomes when having the disease, ORs were

calculated. ORs were calculated separately for Group B and C in comparison to Group A,

respectively. A raw and adjusted form of the ORs are shown. Ratios were adjusted via

multiple logistic regression. The full logistic regression models are shown for the three

primary outcome parameters for Group C with reference to Group A. This was done to

determine the relationship between characteristics of the mother and neonate and adverse

pregnancy outcomes. All other full logistic regressions are displayed in the appendix.

Adjusted ORs were interpreted, when holding all other variables in the models constant.

47

Table 19 Logistic regression: Birth weight > 90th percentile

Parameter Estimate Standard

Error

Wald

Chi-Square

Pr > ChiSq Odds

Ratio

95% Wald

Confidence Limits

Intercept -4.5482 0.2371 368.0424 <.0001

Group C (Insulin

treatment)

0.7344 0.1693 18.8062 <.0001 2.084 1.496 2.905

Age 25 – 29 -0.1786 0.1821 0.9615 0.3268 0.836 0.585 1.195

Age 30 – 34 -0.1625 0.1736 0.8768 0.3491 0.850 0.605 1.194

Age 35 – 39 -0.1465 0.1791 0.6691 0.4134 0.864 0.608 1.227

Age 40+ -0.0270 0.2233 0.0146 0.9037 0.973 0.628 1.508

Female neonate -0.6488 0.0764 72.1023 <.0001 0.523 0.450 0.607

Multiple

gestation

-2.6046 0.7080 13.5347 0.0002 0.074 0.018 0.296

Obesity, Grade 1 0.7207 0.1822 15.6541 <.0001 2.056 1.439 2.938

Obesity, Grade 2 0.4237 0.2322 3.3289 0.0681 1.528 0.969 2.408

Obesity, Grade 3 1.2345 0.1991 38.4534 <.0001 3.437 2.326 5.077

R² 0.00035

Max-rescaled R² 0.0245

Number of

observations

51,918

Dependent variable: Birth weight >90th percentile (yes/no) a Coded as GDM with insulin (Group C) or No GDM (Group A)

For the first primary outcome, birth weight above >90th percentile, GDM with an insulin

treatment significantly increased the likelihood of having the outcome (Table 19). Women

diagnosed with GDM were nearly two times more likely to give birth to a child having a

birth weight above the 90th percentile (OR 2.09 [95% CI 1.50 – 2.91]), independent of other

included variables. Apart from the GDM, only the first and third grade of obesity

significantly increased the probability of having an enlarged neonate. Especially, third grade

obesity was associated with the dependent variable with an OR of 3.43 [95% CI 2.32 - 5.07].

Factors, such as multiple gestation and receiving a female neonate, reduced the likelihood

of the outcome. Age presented in stratified groups showed no significant impact on the

outcome.

When being diagnosed with a GDM and receiving insulin treatment, the probability of a

primary caesarean section is significantly increased for these women (see Table 20). Most

of the variables included, with exception to two age groups, enlarged the probability of

48

having the outcome. Especially women with multiple gestation, an increased age and an

increased BMI were more likely to have a delivery via a primary caesarean section. The

likelihood increased over the grades of obesity. Women with a grade 3 obesity had a higher

chance for a delivery in comparison to women with a grade 1 obesity, when all other

variables are held constant.

Table 20 Logistic regression: Primary caesarean section

Parameter Estimate Standard

Error

Wald

Chi-Square

Pr > ChiSq Odds Ratio 95% Wald

Confidence Limits

Intercept -2.3831 0.0943 638.6645 <.0001

Group C (Insulin

treatment)

0.5313 0.0706 56.6998 <.0001 1.701 1.481 1.953

Age 25 – 29 -0.0384 0.0682 0.3163 0.5738 0.962 0.842 1.100

Age 30 – 34 0.0391 0.0653 0.3597 0.5487 1.040 0.915 1.182

Age 35 – 39 0.3074 0.0663 21.4891 <.0001 1.360 1.194 1.549

Age 40+ 0.6880 0.0771 79.5808 <.0001 1.990 1.711 2.314

Multiple

gestation

1.0499 0.0725 209.4297 <.0001 2.857 2.479 3.294

Obesity, Grade 1 0.4064 0.0745 29.7207 <.0001 1.501 1.297 1.738

Obesity, Grade 2 0.5005 0.0833 36.0680 <.0001 1.650 1.401 1.942

Obesity, Grade 3 0.6949 0.0975 50.7716 <.0001 2.004 1.655 2.426

R² 0.0126

Max-rescaled R² 0.0245

Number of

observations

51,052

Dependent variable: Primary caesarean section (yes/no) a Coded as GDM with Insulin (Group C) or No GDM (Group A)

Neonates, which mothers were diagnosed with gestational diabetes and received insulin, had

a significantly high chance of having a neonatal hypoglycaemia (see Table 21). The

probability in this group is five times as high in comparison to neonates from mothers

without a diagnosis of GDM (OR 5.10 [95% CI 4.09 – 6.36]). Another clear risk factor for

hypoglycaemia in the newborn is multiple gestation. Obesity of the mother was significantly

associated with neonatal hypoglycaemia, if the mother was diagnosed with first or third

grade obesity. None of the age groups included had a significant impact on the outcome.

49

Table 21 Logistic regression: Neonatal hypoglycaemia

Parameter Estimate Standard

Error

Wald

Chi-Square

Pr > ChiSq Odds Ratio 95% Wald

Confidence Limits

Intercept -5.5062 0.1898 841.8323 <.0001

Group C

(Insulin

treatment) a

1.6708 0.1115 224.3830 <.0001 5.317 4.273 6.616

Age 25 – 29 -0.2660 0.1653 2.5881 0.1077 0.766 0.554 1.060

Age 30 – 34 -0.1170 0.1562 0.5612 0.4538 0.890 0.655 1.208

Age 35 – 39 -0.1563 0.1605 0.9476 0.3303 0.855 0.624 1.172

Age 40+ 0.0312 0.1910 0.0267 0.8701 1.032 0.710 1.500

Female

neonate

-0.2557 0.0622 16.9126 <.0001 0.774 0.685 0.875

Multiple

gestation

1.9022 0.0861 488.3425 <.0001 6.701 5.661 7.932

Obesity,

Grade 1

0.3804 0.1706 4.9737 0.0257 1.463 1.047 2.043

Obesity,

Grade 2

0.2930 0.1971 2.2109 0.1370 1.340 0.911 1.972

Obesity,

Grade 3

0.6885 0.1969 12.2268 0.0005 1.991 1.353 2.928

R² 0.0108

Max-rescaled

R²

0.0580

Number of

observations

51,918

Dependent variable: Neonatal hypoglycaemia (yes/no) a Coded as GDM with Insulin (Group C) or No GDM (Group A)

In the following tables, raw ORs were compared to the adjusted ORs for Group B and C and

all outcome parameters.

For the primary outcomes, the adjusted odds ratios were slightly lower in comparison to the

raw odds ratios (see Table 22). Birth weight above 90th percentile became not significant by

adjusting it for independent variables (OR 1.13 [95% CI 0.93 – 1.37]). The two other primary

outcomes were significantly positively associated with the GDM. The chance of the

likelihood to give birth to a child via a primary caesarean section was increased by 16%, and

the odds of clinical neonatal hypoglycaemia more than doubled, when the mother was

diagnosed with GDM. Secondary outcomes, premature delivery, intensive neonatal care,

50

hyperbilirubinemia and preeclampsia were significantly positively associated with mother

diagnosed with gestational diabetes mellitus after adjustment for confounders. Shoulder

dystocia was not significantly associated with GDM, before and after the adjustment of

chosen independent variables.

Table 22 Odds ratios for primary and secondary outcomes: Group B

GROUP A GROUP B

Primary outcome

Raw Odds Ratio

(95% CI)

Adjusted Odds Ratio

(95% CI)

Birth weight >90th percentile a Ref 1.20 (1.00 - 1.45) 1.13 (0.93 - 1.37)

Primary caesarean section b Ref 1.25 (1.18 - 1.34) 1.16 (1.09 - 1.24)

Clinical neonatal

hypoglycaemia a

Ref 2.53 (2.24 - 2.87) 2.38 (2.10 - 2.70)

Secondary outcome

Premature delivery (before 37

wk) a

Ref 1.24 (1.17 - 1.37) 1.15 (1.04 – 1.29)

Shoulder dystocia or birth

injury a

Ref 0.96 (0.69 - 1.34) 0.92 (0.66 - 1.30)

Intensive neonatal care a Ref 1.22 (1.13 - 1.26) 1.17 (1.08 - 1.26)

Hyperbilirubinemia a Ref 1.27 (1.14 - 1.41) 1.22 (1.10 - 1.36)

Preeclampsia b Ref 1.45 (1.36 - 1.61) 1.32 (1.22 - 1.44)

a: age of the mother, sex of the neonate, multiple gestation, 1st, 2nd, and 3rd obesity of the mother were

included into the logit model

b: age of the mother, multiple gestation, 1st, 2nd, and 3rd obesity of the mother were included into the

logit model

In Table 23 the odds ratio of primary and secondary outcomes for Group C are presented.

Also, as for Group B, Group A, served as reference group for the calculation of ORs.

Variables used for adjustment remained the same. Comparable to Table 22 all adjusted odds

ratios were lower in comparison to the raw ratios. All three primary outcome parameters

were significant, before and after adjustment. The odds of a pregnant women with a GDM

treated with insulin having a newborn with increased weight more than doubled in

comparison to the general pregnant population (OR 2.08 [95% CI 1.50 – 2.90]). The

association was even stronger for the outcome of clinical neonatal hypoglycaemia, here the

odds ratio was 5.32 [95% CI 4.27 -6.62]. Secondary outcomes were not as clearly positively

associated with the exposure of GDM and insulin treatment. Only two of the five secondary

outcomes showed a significant association. The likelihood of intensive neonatal care and

preeclampsia increased significantly when having a GDM treated with insulin, with ORs of

1.25 [95% CI 1.04 – 1.50] and 1.51 [95% CI 1.23 – 1.85], respectively.

51

Table 23 Odds ratios for primary and secondary outcomes: Group C

GROUP A GROUP C

Primary outcome

Raw Odds Ratio

(95% CI)

Adjusted Odds Ratio

(95% CI)

Birth weight >90th percentile a Ref 2.59 (1.88 - 3.57) 2.08 (1.50 – 2.90)

Primary caesarean section b Ref 1.99 (1.74 - 2.28) 1.70 (1.48 – 1.95)

Clinical neonatal

hypoglycaemia a

Ref 5.45 (4.42 - 6.70) 5.32 (4.27 – 6.62)

Secondary outcome

Premature delivery (before 37

wk) a

Ref 1.13 (0.88 - 1.45) 1.22 (0.94 – 1.59)

Shoulder dystocia or birth

injury a

Ref 1.32 (0.65 - 2.67) 1.12 (0.55 – 2.30)

Intensive neonatal care a Ref 1.25 (1.07 – 1.54) 1.25 (1.04 – 1.50)

Hyperbilirubinemia a Ref 0.81 (0.60 – 1.10) 0.79 (0.58 – 1.08)

Preeclampsia b Ref 1.98 (1.63 – 2.40) 1.51 (1.23 – 1.85)

a: age of the mother, sex of the neonate, multiple gestation, 1st, 2nd, and 3rd obesity of the mother were

included into the logit model

b: age of the mother, multiple gestation, 1st, 2nd, and 3rd obesity of the mother were included into the logit

model

5.3. Period of follow-up “t2”

In the time after pregnancy, the most relevant numbers of this analysis were the rate of

women who received a follow-up test (75g OGTT) and the prevalence of diabetes mellitus

type 2, both in the year after delivery. Table 24 shows the number of women, which received

a 75g OGTT test in the year after delivery. Group C had the highest share, with 49.32%

women tested. The rate in Group B was lower. Only one third of women with a GDM

diagnosis, without insulin treatment, got tested in the year after delivery (n=2,160; 29.81%).

In the two groups of women with a diagnosis, most of the test were utilized within the first

18 weeks after pregnancy. In Group B (66.37% of all tests) and in Group C (76.37% of all

tests). In contrast, in Group A of all women who got tested less than 40% were tested after

18 weeks after pregnancy. Rates at 12 weeks, the time recommended in the guidelines, after

pregnancy for Group A, B and C were 4.24%, 14.33% and 32.06% of women who got tested

with a 75 OGTT, respectively.

52

Table 24 Follow-up 75g OGTT

Group A

(No GDM)

Group B

(GDM)

Group C

(GDM with insulin)

Number 49,645 7,245 1,407

OGTT within 1-year

after pregnancy 8,024 (16.16%) 2160 (29.81%) 694 (49.32%)

OGTT, between

0-6 weeks 1,122 (14.01%) 238 (11.02%) 75 (10.81%)

7-12 weeks 982 (12.26%) 706 (32.70%) 277 (39.91%)

13-18 weeks 976 (12.18%) 489 (22.65%) 178 (25.65%)

19-24 weeks 1031 (12.87%) 193 (8.94%) 39 (5.62%)

25 weeks until 1 year 3900 (48.68%) 533 (24.69%) 125 (18.01%)

Interestingly, the percentage of doctor specialty group applying the test, were largely

different within the three groups (see Table 25). Whereas, in two-third of all cases, the

diabetologists applied the test in Group C, only 17% of all tests were applied by the same

specialty group in Group A. Over the three study groups, only three specialty groups were

involved into the process of testing after pregnancy. The diabetologist, the general

practitioner and the laboratory physicians. The specialty group of gynaecologists largely

disappeared from the process of follow-up testing, with rates below 1.5% in all study groups.

This is even more surprising, when considering, that gynaecologist carried out a large share

of screening tests of GDM.

53

Table 25 75g OGTT by doctor specialty groups

Group A

(No GDM)

Group B

(GDM)

Group C

(GDM with insulin)

Number 49,645 7,245 1,407

Total number

tested 8,024 (16.16%) 2,160 (32.78%) 694 (49.32%)

Diabetology/GP with

additional training 1,386 (17.27%) 1,003 (46.44%) 439 (63.26%)

GP without

additional training 2,183 (27.21%) 365 (16.90%) 101 (14.55%)

Laboratory medicine 2,672 (33.30%) 473 (21.90%) 89 (12.82%)

Angiology 4 (0.05%) 21 (0.97%) 6 (0.86%)

Nephrologie 47 (0.59%) 15 (0.69%) 5 (0.72%)

Gynecology 116 (1.45%) 23 (1.06%) 3 (0.43%)

Others 375 (4.67%) 74 (3.43%) 15 (2.16%)

Not defined 1,241 (15.47%) 186 (8.61%) 36 (5.19%)

The transition rate from a GDM diagnosis to diabetes mellitus type 2 was of interest.

Numbers are displayed in Figure 9. Comparable to the test applied, women in Group C had

the highest 1-year rate of diabetes mellitus type 2, with a rate of 18.41%. The rate is four

times higher in comparison to Group B, with a rate of 6.21%. In both groups diagnosed with

a GDM, more than 75% of the diagnosis were made in the first two quarters after pregnancy.

This might reflect, the time, when the follow-up tests were applied. Here, most of the tests

were performed in the first six months after pregnancy in these two groups (see Figure 9).

Figure 9 Cumulative 1-year rate of diabetes mellitus type 2

0,1% 0,2% 0,3% 0,5%2,0%

4,8% 5,5% 6,2%5,1%

13,8%15,8%

18,4%

0%

5%

10%

15%

20%

1st quarter 2nd quarter 3rd quarter 4th quarter

Per

cen

tag

e o

f g

rou

p

Time after pregnancy

Cummulative 1-year rate of diabetes mellitus type 2

Group A (n=49,645) Group B (n=7,245) Group C (n=1,407)

54

6. Discussion

Gestational diabetes mellitus is a serious health condition appearing during pregnancy,

which affects the short and long-term health of mothers and the new born. In 2011 a new

guideline was implemented to identify the GDM in pregnant women and lower the burden

for mother and child by effective treatment, e.g. change of lifestyle factors or the use of

pharmacological agents. Empirical evidence is missing on how the guideline was

implemented into the German system, including the screening, treatment, and follow-up

after pregnancy. Most important, no population wide data is available on the health status of

mother and child, measured by adverse events during hospitalization. Describing the

implementation and the occurrence of adverse events is needed to clarify the current burden

of disease.

6.1. Comparing of findings with current research

The most striking result is that many adverse events are clearly associated with the GDM.

Especially, the three primary outcomes (Birth weight >90th percentile, primary caesarean

section, clinical neonatal hypoglycaemia), are more likely in the GDM groups. When the

GDM was treated with insulin, this association was even more apparent, before and after

adjusting for major confounding factors. It needs to be considered, that not the given insulin,

but the reason insulin was prescribed for these women is associated with the chosen adverse

events. According to the guideline, insulin should be considered by diabetologists, when

50% of the time the hypoglycaemic target values are exceeded (Kleinwechter et al., 2011, p.

45). Therefore, it can be expected that women with a prescription of insulin, have higher

blood glucose levels. Findings would then be in line with the HAPO study, where prove was

given, that an increased level of blood glucose measures is positively associated with major

adverse events, such as primary caesarean section or macrosomia (Metzger et al., 2008). A

French health claim data study, with a comparable methodology, also stratified women with

GDM into “treated” and “non-treated” and came to comparable results for odds ratios of

caesarean section, pre-eclampsia and macrosomia for both groups (Billionnet et al., 2017).

Clinical neonatal hypoglycaemia, however, is in the epidemiological HAPO study not as

clearly associated with increased rates of hypoglycaemia, as is it in this health claims data

analysis. In this study Group B and C had ORs of 2.45 [95% CI 2.15 – 2.79] and 5.10 [95%

55

CI 4.09 – 6.35]. The highest association in the HAPO study for this outcome was an OR of

2.17 [95% CI 1.28 – 3.69] in the second highest maternal blood glucose group defined within

the study (Metzger et al., 2008). Difference between studies could root in the difference of

the methods. In the HAPO study, data about the blood glucose levels have been blinded for

health care workers. Blinded data is not available in real-world studies, where the diagnosis

is especially known, when women receive pharmacological agents. If a diagnosis/treatment

is known, this has an impact on assessing the disease status and in the decision making of

health care workers (Day & Altman, 2000). Due to this, neonates from mothers with GDM

might have been observed more closely after birth than neonates from mothers without,

resulting in higher rates of diagnoses.

A further interesting result is the confirmation that GDM or an increased blood glucose level

and obesity are both independent risk factors for the occurrence of macrosomia and the

primary caesarean section. In an outline report of the HAPO study the same relationship was

described. Odds ratios for birthweights >90th and primary caesarean section were nearly

linearly associated with BMI. There was no significant relationship for an increased BMI

and the occurrence of clinical neonatal hypoglycaemia in the HAPO data (McIntyre et al.,

2010). However, this study showed an increased likelihood, when mothers were diagnosed

with third grade obesity.

Aside from adverse events reported during hospitalization for delivery, some interesting

results were observed within the health care service research aspect of this analysis.

An interesting finding was that not only most women were tested for the GDM at least once,

but that most of those were tested at the time of gestation recommended by the guideline, so

between 24th and 28th week of gestation (Kleinwechter et al., 2011, p. 22). The 75g OGTT

is, however, used over the whole gestational period. This might reflect a risk/symptom-based

screening by doctors.

Based on the screening and the different levels of validity, the prevalence of gestational

diabetes mellitus in Germany is between 9.87 – 12,73%. A similar prevalence was estimated

with nationwide outpatient database, with an estimate of 13.1% (Melchior et al., 2017).

Further, the prevalence is comparable to outcomes of a meta-analysis, where prevalence for

the IADPSG criteria was described as 14.1% [95% CI 8.9 – 21.5]. Ten studies were

considered for this estimate (Eades et al., 2017).

56

The insulin rate (14%) of this study is considerably lower than the rate of the GestDiab-

registry, reporting a rate of nearly 40%. However, it needs to be considered that this data is

only based on 28 diabetology specialist practices (Adamczewski et al., 2016). In this

administrative data report, it was shown that, only half of the women with a GDM diagnosis

got referred to a GP office specialized in diabetology. When a diagnosis was coded by a

diabetologist, the rate was still lower with 27%, so considerably lower than the rate reported

in the GestDiab. The administrative database report from France reported an insulin rate of

28.1%. Instead of population wide-screening, like in Germany, in France a risk-based

screening is applied. Resulting in a lower prevalence and consistently more women that are

in need of a pharmacological treatment (Billionnet et al., 2017). In the MFMU-Trial a

percentage of 7.4% of mothers with GDM received insulin (Landon et al., 2009). The

algorithm proposed in the MFMU-Trial was implemented into the German S3 guideline

(Kleinwechter et al., 2011, p. 30) . The insulin rate could still be characterized as increased

in comparison to the rate reported in the clinical trial. Data from other countries with similar

screening algorithms, is needed to describe a natural rate of women needed for an insulin

prescription.

Collaboration between the two most important specialty groups, gynaecologists and

diabetologists, is not developed, as only less than half of the population receives a diagnosis

from the two specialist groups. Gynaecologists, the specialty group establishing the

diagnosis most of the time, might have accumulated knowledge and expertise in the

treatment of a milder form of gestational diabetes mellitus. Those doctors might only refer

pregnant women to a diabetologist, if blood glucose targets are not met within the first weeks

after establishing the diagnosis. This would explain the higher number of mothers receiving

insulin, when a diagnosis has been recognized by a diabetologist. A qualitative report by

Diehl et al. is not in support of the drawn hypothesises, where all participating

gynaecologists cooperated with diabetologists. Further, the report stated that no time was

available for gynaecologists to give women nutritional counselling (Diehl et al., 2016).

However, the report was conducted in the year after implementation of the screening

algorithm in 2013. Gynaecologists might have adopted and offer nutrition counselling now.

The last observation period, the time after pregnancy, revealed that utilization of follow-up

tests is low. Even in the group treated with insulin only half of the population utilized the

test within the following year after delivery. In general, the rate of follow-up is comparable

to research from the US, were a rate of 24% after 6 months postpartum was estimated

57

(McCloskey et al., 2014). Interestingly, the gynaecologists, involved in testing for the GDM

during pregnancy is not involved in testing for diabetes mellitus postpartum. When also

considering that 50% of the established diagnoses do not get referred to a diabetologist and

only receive a diagnosis by a gynaecologist, this case gets even stronger. In the maternity

directives (Mutterschafts-Richtlinien), published by the G-BA, all services that can be

claimed by insurees during and after pregnancy are defined. Maternity directives were

implemented to reduce the health burden for mother and child. For example, the screening

for GDM is also defined in the directives. Interestingly, a postpartum follow-up care is

already established. Between 6-8 weeks postpartum different test and counselling is

recommended in this catalogue, including findings from a gynaecologist. Counterintuitively,

no postpartum follow-up test for diabetes is defined in this catalogue (G-BA, 2016). The

already established postpartum care program could be an opportunity to increase numbers

of women with a GDM to receive testing after pregnancy by a gynaecologist. Gynaecologists

are already well equipped and familiar with the 75g OGTT test, as 75% of all women in need

of this test during pregnancy, received it within an office of a gynaecologist.

6.2. Advantageous of the study

This analysis is the first nationwide study focusing on outcome parameters of mother and

child in women with gestational diabetes mellitus in Germany. The HAPO study focused on

15 centres in nine countries and included more than 20,000 women (Metzger et al., 2008).

In this analysis, nearly 60,000 births have been included for analysis with an extend focus

on the real-world care situation in Germany. All hospitals and doctors involved into the

provision of care were included into the perspective, in comparison to a small number of

excellent centres in the epidemiological HAPO study.

The HAPO study showed, that not just the health of the mother but the health of the new

born is affected by GDM. Therefore, health care service research needs to take in the health

of the child to evaluate all aspects of the disease. This aspect is missing in the only German

health care service research study focusing on adverse events of a GDM (Tamayo et al.,

2016). The linkage of mother and new orn, which was possible in 79% of all identified

pregnancies, made it possible to evaluate the incidence of adverse events of the infant. The

number of mother linked to their baby is comparable to the initial study using the technique

58

(Garbe et al., 2011). Mother-baby pairs made it possible to investigate the occurrence of

adverse events of the neonate, such as clinical neonatal hypoglycaemia.

The second advantage of this analysis were the many methodological steps taken to account

for misdiagnosis of a GDM, hence increasing the robustness of the diagnosis. First, mothers

were excluded without one out of the two refundable tests or not having a prescription for

blood glucose stripes. Therefore, all women in the next step of the analysis were at least

knowingly tested once for the GDM. Previously, it was considered to exclude also women

with a diagnosis, but without a confirmatory test. However, the guideline describes that a

50g GCT with an outcome of above >200mg/dl after 1 hour is already enough to establish a

GDM diagnosis (G-BA, 2012). To exclude women after receiving a 50g GCT without a 75g

OGTT might have excluded women with a clear diagnosis of GDM. In a further step, women

were excluded from the GDM groups, if they had a coded diabetes mellitus or received

insulin one year prior to pregnancy. By this, coding errors, coding diabetes mellitus type 1/2

falsely as GDM, were excluded from those groups and the prevalence estimate.

In this analysis, data from all important sectors, that are involved in a) establishing the

diagnosis, b) in the process of treatment c) the delivery of the newborn d) follow-up of the

mother with GDM are available. The sectors involved are pharmacies and the in- outpatient

sector. An analysis that only focus on adverse event of mother and child during the

hospitalisation of delivery, could underestimate the prevalence of GDM and overestimate

rates of adverse events. The screening and therefore the diagnosis are conducted in the

outpatient sector. Not all women, especially those with a mild gestational diabetes mellitus,

might have a coded GDM diagnosis during their time of hospitalization, as the diagnosis

seems irrelevant for the treatment and process. Women with a GDM treated with insulin

have will have a higher likelihood of a coded GDM diagnosis in the inpatient sector. By

connecting diagnosis from the outpatient sector with the adverse event coded in the inpatient

sector, a truer estimate of prevalence and rates of adverse events is possible. The Aqua

institute, a private institution responsible for the external quality assurance for the hospital

sector, estimated a prevalence of 4.5% for the gestational diabetes mellitus in the year of

2014. The institute received 690.000 data records from over 700 hospitals in Germany for

the year 2014 (AQUA-Institut, 2015, p. 99). Melchior et al. estimated a prevalence of 13.2%

based on outpatient record data for the year of 2014, as well (Melchior et al., 2017). These

two estimates show the discrepancy between the two datasets and the need for cross-sector

analysis.

59

6.3. Limitations of the study

There are many considerations, when working with health claims data. It needs to be

considered that health claims data are only reused for scientific purpose, whereas the first

intention of collection of the data is designed for remuneration purposes. Data may be biased

due to incentives in reimbursement (Ohlmeier et al., 2014). A way in which data may be

biased is the so called upcoding of patients. Hereby, patients are intentionally miscoded, so

they seem to be sicker than they are, to generate higher reimbursement for the hospital.

Evidence is available that this illegal upcoding is apparent in neonatology. One of the main

factors needed to generate the DRG for the admission of neonates is the birthweight. A

falsely stated birth weight is non-verifiable by payers. Change in the DRG can conclude in

additional payment of more than 15,000 €, so a monetary incentive for hospital is given to

conduct falsely stated birthweight of neonates. Between the years of 2006 and 2011 it is

assumed that nearly 12,000 births were up-coded within the German DRG system (Jürges &

Köberlein, 2015). The neonatal birthweight is one of the primary outcomes of the study and

could be affected by this. However, birthweight thresholds only exist in the DRG catalogue

for weights below 2,499g, while those DRG are specifically for pre-term deliveries (InEK

GmbH, 2016, p. 59). Thus, there is no financial incentives to miscode birthweights above

this threshold and weights above the >4,000g, i.e. the 90th percentile. However, there is also

no incentive to code birth weights above 2,500g at all. Still, when comparing the rate to

nationwide data from the federal statistic office, the rate seems plausible, 1.6% of all mothers

within the year of 2013 had a new born with a birthweight above 4,500g (Robert Koch-

Institut, 2018).

Apart from upcoding, underreporting or non-reporting of conditions is a limitation of health

claims data. Conditions are unintentionally not coded, when they are not relevant for

reimbursement purposes. An analysis in the U.S. compared the reporting of two serious

neurological conditions, epilepsy and multiple sclerosis, in pregnant women pre-delivery to

the reporting at delivery. Women diagnosed with epilepsy before the delivery were only

coded with the condition in 70% of the cases in the hospital discharge data. For multiple

sclerosis this rate was at 60% (MacDonald, Hernán, McElrath, & Hernández-Díaz, 2018).

Even though that data comes from another health care system, with other structures, it shows

that even serious conditions are not well reported in hospital discharge data. The conducted

study relied in many cases on adverse events reported in the hospital only. All adverse events

60

relied on hospital data only, without the possibility of cross-verification by other data.

Condition in the report could therefore be underestimated.

The analysis estimated that 82% of all women received at least one test, 50g GCT or 75g

OGTT, to screen for a GDM. Unfortunately, it cannot be concluded that the other 18% did

not conduct any of the tests. Disagreement about the screening algorithm exists between the

medical societies and the G-BA. The G-BA prefers the existing two-step population

screening, whereas the new guideline recommends to only conduct the 75g OGTT (DGG &

DGGG, 2018, p. 19; G-BA, 2012, p. 11). Already after implementation of the screening

algorithm, some doctors only recommended the 75g OGTT test, without the 50g GCT

beforehand. The OGTT without a previous GCT cannot be reimbursed by SHI’s and would

need to be self-paid by pregnant women (Diehl et al., 2016). This leads to the implication,

that an uncountable number of women with a true diagnosis, established by a self-paid 75g

OGTT test, were excluded from the further analysis. To account for this problem, another

tool to verify a GDM diagnosis was used. Women with a diagnosis and a prescription for

blood-glucose stripes were additionally considered as having a valid diagnosis. However,

there were still women with a diagnosis, without a test or blood glucose test stripes

prescribed.

The transition from a GDM to diabetes mellitus type 2 has also some limitations. Women in

the group with GDM and an insulin treatment were much more likely to get tested for a

gestational diabetes mellitus and therefore a diagnosis might have been established more

often. In general, the validity of the transition rate from a GDM to diabetes mellitus type 2

is not valid as only one small proportion of women got tested after delivery.

Two additional groups of women were excluded from the study. Firstly, only deliveries in

hospital were included into the study, excluding home-childbirth and other forms of

outpatient delivery. It needs to be considered that less than 2% of all mothers had a delivery

outside of a hospital in 2010 (Albrecht, Loos, Sander, Schliwen, & Wolfschütz, 2012).

Secondly, the large proportion of 21% of women without a possible linkage to their newborn

in the SHI dataset. Non-linkage could be associated with determinants of health (e.g. social,

environmental and individual criteria) impacting health of mother and child. Especially, the

exclusion of the second group might diminish the generalizability of the study.

Future analyses should focus on the long-term effects for mother and neonate of a GDM.

The current report was only designed to detect health outcomes during hospitalization. A

61

continuing follow up using health claims data could further be established to describe the

long-term burden of the disease for the mother and the child. Only one study looked at the

long-term effect for children. (Clausen et al., 2008). Data may be outdated, as data existed

prior to 2008, and is restricted to a Danish population. Also, interventions need to be

established and researched to increase rate of women utilizing follow-up tests for diabetes

type 2 mellitus in Germany. One aspect could be the involvement of the gynecologists to

apply tests during the already established follow-up in the maternity directive. A third

question that should be targeted in the future is research about the communication between

doctor specialty groups to improve diagnostic, treatment and follow-up of a GDM.

7. Conclusion

This study set out to determine the burden of a gestational diabetes mellitus in Germany for

mother and child. Further, critical aspect in the provision of care described in the S3-

guideline were evaluated. The investigation has shown that most of the pregnant women

receive general screening. However, not all women are referred to a diabetologist, when

diagnosed with a GDM. The most obvious finding, that has been analysed by previous

studies, were an increased risk for several adverse events in women with GDM in Germany.

With an insulin treatment, indicating a higher rate of hypoglycaemia, the association

increased. After pregnancy, the minority of women with a GDM receives follow-up

monitoring. Long-term follow-up studies are needed to measure the impact of GDM and

adverse pregnancy events on the health of the mother and child in years after pregnancy.

Moreover, intervention strategies are needed to increase the participation in monitoring of

the disease in the future.

62

8. Excurse: Validation of the obstetric comorbidity index

The excurse describes and discusses the results of the validation of the obstetric comorbidity

index briefly. Finally, an explanation for the exclusion of the index will be given.

8.1. Results of the validation

From the 88,632, 78,730 women were eligible for the validation of the obstetric comorbidity

index. Therefore, 9,632 women were excluded from the validation, as a coded end-organ

damage diagnosis before the hospitalization of delivery was present. The Elixhauser Score

and the OCI were calculated for all women eligible.

Table 26 Distribution of OCI and end-organ damage

Group 0 1 2 3 4 5 6 7 8 9 10+

OCI, n 44328 22547 8154 2138 757 444 199 101 30 17 15

OCI, % 56.30 28.64 10.36

2.72 0.96 0.56 0.25 0.13 0.04 0.02 0.02

End-organ damage, n

417 266 139 67 260 156 91 37 10 12 10

End-organ damage, %

0.9 1.2 1.7 3.1 34.3 35.1 45.7 36.7 33.3 70.6 66.7

χ² (10) = 11114.58; p <0.001 Cramer's V: 0.375

Table 26 present the distribution of the obstetric comorbidity index over the pregnant

population. Overall, the distribution of the OCI is skewed right, having large percentages in

categories with small values, and low percentages in high values. Most of women were

classified with a comorbidity score of 0 (n= 44,328; 56.30%). More than 95% of all women

had a score equal or lower than the first three categories.

The incidence of end-organ damage is significantly correlated with an increased OCI (X²

(11) = 11114.58; p: <0.001). Cramer’s V revealed a moderate correlation between the two

variables (Cramer’s V: 0.375). The incidence of end-organ damage remained low, within the

first three values and increased sharply in the fourth value of the OCI. From 757 women

with a score of four, 260 (34.3%) were diagnosed with an end-organ damage during the

63

hospitalization of delivery up to 30 days afterwards. Over the next groups, incidence of end-

organ damage remained stable. A further increase was noticeable for the scores of 8 -10+,

where the share increased to 70% of the population with an end-organ damage. It should be

noticed that sample sizes were small in the higher scores.

In comparison to the OCI, the distribution of the Elixhauser score was different (see Table

27). No clear distribution was observed. Most of women were classified with a score 0

(n=42,946; 54.5%). The second and third largest score group were women a score of -2

(n=20,078; 25.5%) and 3 (n=6,694; 8.5%), respectively. All other women were distributed

equally among the rest of the score groups.

Table 27 Distribution of the Elixhauser Score and end-organ damage

Group -2 -1 0 1 2 3 4 5 6 7 8

Elixhauer, n

20078 1215 42946 1405 767 6694 977 1559 1008 330 1751

Elixhauser, %

25.5 1.5 54.5 1.8 1.0 8.5 1.2 2.0 1.3 0.4 2.1

End-organ damage, n

401 34 614 46 22 140 33 38 25 17 95

End-organ damage, %

2.0 2.8 1.4 3.2 2.8 2.1 3.4 2.4 2.5 5.2 5.7

χ² (10) = 231.85; p <0.001 Cramer's V: 0.054

The incidence of end-organ damage within the score-groups, did not increase in women with

higher scores. Even though there was a significant interaction, between an increased score

and the likelihood of having an end-organ damage [X² (10) = 231.85; p <0.001], correlation

between the two variables was low (Cramer’s V: 0.054).

64

Table 28 Frequency of comorbidities of the obstetric comorbidity index

Condition Weight in OCI Frequency Percentage

Alcohol abuse 1 3 0.00%

Asthma 1 244 0.31%

Cardiac Valvular Disease 2 43 0.05%

Chronic Congestive Heart

Failure

5 0 0.00%

Chronic Ischemic Heart

Disease

3 5 0.01%

Chronic Renal Disease 1 231 0.29%

Congenital Heart Disease 4 540 0.69%

Drug Abuse 2 31 0.04%

Gestational hypertension 1 1,098 1.38%

Human

Immunodeficiency Virus

2 23 0.03%

Mild/Unspecified Pre-

Eclampsia

2 914 1.16%

Multiple Gestation 2 1,576 2.00%

Placenta Previa 2 377 0.48%

Pre-Exisiting Diabetes

Mellitus

1 1,058 1.34%

Pre-Existing

Hypertension

1 578 0.73%

Previous Cesarean

Delivery

1 8,010 10.17%

Pulmonary Hypertension 4 1 0.00%

Severe Pre-Eclampsia 5 341 0.43%

Sickle Cell Disease 3 28 0.04%

Systemic Lupus

Erythematosus

2 12 0.02%

Age groups,

Age at delivery 35-39 1 22,333 28.37%

Age at delivery 40-44 2 4,311 5.48%

Age at delivery >44 3 296 0.38%

Table 28 presents an overview of the frequency of the comorbidities included into the

obstetric comorbidity index. From the chosen 20 conditions, only five were occurring in

more than 1% of all women. From these five conditions only, previous caesarean delivery

was reported more frequently.

65

Figure 10 ROC curve: OCI and Elixhauser Score

The results of the ROC curve are presented in Figure 12. The ROC compares the

discriminatory power of the Elixhauser Score and the OCI in predicting maternal end-organ

damage. It is apparent from this figure that the OCI had a larger area under the ROC curve.

Therefore, the OCI had a higher discriminatory power, predicting maternal end-organ

damage with a higher likelihood than the Elixhauser Score.

8.2. Discussion of the validation

The OCI score was calculated at the delivery of hospitalization and included 20

comorbidities and three age groups. AUC values for the OCI were comparable to other

research. In the initial study of the index an AUC of 0.66 [95% CI 0.65 – 0.67] and for the

first ICD-10 translation by Metcalfe et al, an AUC of 0.70 [95% CI 0.60 – 0.80] was

calculated (Bateman et al., 2013; Metcalfe et al., 2015). Based on these findings, the score

is a valid measure to adjust for comorbidities in an obstetric population.

66

Still the score was excluded from the initial study for several reasons. First, the score did not

show a high prevalence in many of the chosen comorbidities. Further, those conditions,

where a higher prevalence was observed, were either outcome parameters, such as pre-

eclampsia or due to their importance already included separately in the model, such as

multiple gestation or age. Secondly, it was decided to interpret, and not only adjust for

associations between dependent and independent variables. However, a comorbidity index

is an abstract concept, with no ability to generalize findings or compare them to other

research conducted, such as the HAPO study. Finally, only confounders used in other

epidemiological GDM studies, applicable in health care service research, were included into

the models. These were obesity, age and multiple gestation (Cozzolino et al., 2017; McIntyre

et al., 2010).

67

9. References

ACOG. (2016). Practice Bulletin No. 173: Fetal Macrosomia. Obstetrics and Gynecology,

128(5), e195–e209. https://doi.org/10.1097/AOG.0000000000001767

ADA. (2017). Standard of medical care in diabetes - 2017. Diabetes Care, 40 (sup

1)(January), s4–s128. https://doi.org/10.2337/dc17-S003

Adamczewski, H., Weber, D., Faber-Heinemann, G., Heinemann, L., & Kaltheuner, M.

(2016). Einfluss der Gestationsdiabetes-Leitlinie der DDG auf die Versorgungsrealität:

Analysen des Register GestDiab. Diabetologie Und Stoffwechsel, 11(5), 341–349.

https://doi.org/10.1055/s-0042-110485

Al-Jameil. (2013). A Brief Overview of Preeclampsia. Journal of Clinical Medicine

Research, 6(1), 1–7. https://doi.org/10.4021/jocmr1682w

Albrecht, M., Loos, S., Sander, M., Schliwen, A., & Wolfschütz, A. (2012). Versorgungs-

und Vergütungssituation in der außerklinischen Hebammenhilfe - Ergebnisbericht für

das Bundesministerium für Gesundheit; IGES Institut GmbH. Berlin.

Aoyama, K., D’Souza, R., Inada, E., Lapinsky, S. E., & Fowler, R. A. (2017). Measurement

properties of comorbidity indices in maternal health research: A systematic review.

BMC Pregnancy and Childbirth, 17(1), 1–8. https://doi.org/10.1186/s12884-017-1558-

3

AQUA-Institut. (2015). Bundesauswertung zum Erfassungsjahr 2014 - 16/1 Geburtshilfe

(Qualitätsindikatoren 16/1). Retrieved from

https://www.sqg.de/downloads/Bundesauswertungen/2014/bu_Gesamt_16N1-

GEBH_2014.pdf

Bateman, B. T., Mhyre, J. M., Hernandez-Diaz, S., Huybrechts, K. F., Fischer, M. A.,

Creanga, A. A., … Gagne, J. J. (2013). Development of a comorbidity index for use in

obstetric patients. Obstetrics and Gynecology, 122(5), 957–65.

https://doi.org/10.1097/AOG.0b013e3182a603bb

Ben-Haroush, A., Yogev, Y., & Hod, M. (2004). Epidemiology of gestational diabetes

mellitus and its association with Type 2 diabetes. Diabetic Medicine : A Journal of the

British Diabetic Association, 21(2), 103–13. https://doi.org/10.1046/j.1464–

68

5491.2003.00985.x

Benhalima, K., Mathieu, C., Damm, P., Van Assche, A., Devlieger, R., Desoye, G., …

Dunne, F. (2016). Erratum to: A proposal for the use of uniform diagnostic criteria for

gestational diabetes in Europe: an opinion paper by the European Board & College of

Obstetrics and Gynaecology (EBCOG) (Diabetologia, 10.1007/s00125-015-3615-7).

Diabetologia, 59(8), 1791. https://doi.org/10.1007/s00125-016-3994-4

Benhalima, K., Van Crombrugge, P., Moyson, C., Verhaeghe, J., Vandeginste, S.,

Verlaenen, H., … Mathieu, C. (2018). The sensitivity and specificity of the glucose

challenge test in a universal two-step screening strategy for gestational diabetes mellitus

using the 2013 World Health Organization criteria. Diabetes Care, 41(7), e111–e112.

https://doi.org/10.2337/dc18-0556

Bergsjø, P., Denman, D. W., Hoffman, H. J., & Meirik, O. (1990). Duration Of Human

Singleton Pregnancy: A Population-based Study. Acta Obstetricia et Gynecologica

Scandinavica, 69(3), 197–207. https://doi.org/10.3109/00016349009028681

Billionnet, C., Mitanchez, D., Weill, A., Nizard, J., Alla, F., Hartemann, A., & Jacqueminet,

S. (2017). Gestational diabetes and adverse perinatal outcomes from 716,152 births in

France in 2012. Diabetologia. https://doi.org/10.1007/s00125-017-4206-6

Brown, J., Alwan, N. A., West, J., Brown, S., McKinlay, C. J., Farrar, D., & Crowther, C.

A. (2017). Lifestyle interventions for the treatment of women with gestational diabetes.

Cochrane Database of Systematic Reviews.

https://doi.org/10.1002/14651858.CD011970.pub2

Brown, J., Grzeskowiak, L., Williamson, K., Downie, M. R., & Crowther, C. A. (2016).

Insulin for the treatment of women with gestational diabetes. Cochrane Database of

Systematic Reviews, 2016(1). https://doi.org/10.1002/14651858.CD012037

Buchanan, T. A., & Xiang, A. H. (2005). Gestational diabetes mellitus. The Journal of

Clinical Investigation, 115(3), 485–91. https://doi.org/10.1172/JCI24531

Catalano, P. M. (2014). Trying to understand gestational diabetes. Diabetic Medicine : A

Journal of the British Diabetic Association, 31(3), 273–81.

https://doi.org/10.1111/dme.12381

Catalano, P. M., Avallone, D. A., Drago, N. M., & Amini, S. B. (1993). Reproducibility of

69

the oral glucose tolerance test in pregnant women. American Journal of Obstetrics and

Gynecology, 169(4), 874–881. https://doi.org/10.1016/0002-9378(93)90019-F

Catalano, P. M., Huston, L., Amini, S. B., & Kalhan, S. C. (1999). Longitudinal changes in

glucose metabolism during pregnancy in obese women with normal glucose tolerance

and gestational diabetes mellitus. American Journal of Obstetrics and Gynecology,

180(4), 903–16. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/10203659

Clark, H. D., Graham, I. D., Karovitch, A., & Keely, E. J. (2009). Do postal reminders

increase postpartum screening of diabetes mellitus in women with gestational diabetes

mellitus? A randomized controlled trial. American Journal of Obstetrics and

Gynecology, 200(6), 634.e1-634.e7. https://doi.org/10.1016/j.ajog.2009.01.003

Clausen, T., Mathiesesn, E., Hansen, T., Pedersen, O., Jensen, D., Launborg, D., & Damm,

P. (2008). High Prevalence of Type 2 Diabetes and Pre-Diabetes in Adult Offspring of

Women With Gestational Diabetes The role of intrauterine hyperglycemia. Diabetes

Care, 31(2), 340–346. https://doi.org/10.2337/dc07-1596.Additional

Cozzolino, M., Serena, C., Maggio, L., Rambaldi, M. P., Simeone, S., Mello, G., …

Mecacci, F. (2017). Analysis of the main risk factors for gestational diabetes diagnosed

with International Association of Diabetes and Pregnancy Study Groups (IADPSG)

criteria in multiple pregnancies. Journal of Endocrinological Investigation, 40(9), 937–

943. https://doi.org/10.1007/s40618-017-0646-6

Crowther, C. A., Hiller, J. E., Moss, J. R., McPhee, A. J., Jeffries, W. S., & Robinson, J. S.

(2005). Effect of Treatment of Gestational Diabetes Mellitus on Pregnancy Outcomes.

New England Journal of Medicine, 352(24), 2477–2486.

https://doi.org/10.1056/NEJMoa042973

Dabelea, D., Snell-Bergeon, J. K., Hartsfield, C. L., Bischoff, K. J., Hamman, R. F.,

McDuffie, R. S., & Kaiser Permanente of Colorado GDM Screening Program. (2005).

Increasing prevalence of gestational diabetes mellitus (GDM) over time and by birth

cohort: Kaiser Permanente of Colorado GDM Screening Program. Diabetes Care,

28(3), 579–84. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/15735191

Day, S. J., & Altman, D. G. (2000). Statistics notes: blinding in clinical trials and other

studies. BMJ (Clinical Research Ed.), 321(7259), 504.

https://doi.org/10.1136/BMJ.321.7259.504

70

DDG. (2013). Richtlinien zur Anerkennung als Diabetologe / in DDG, (September).

Retrieved from https://www.deutsche-diabetes-

gesellschaft.de/fileadmin/Redakteur/Weiterbildung/Diabetologe_DDG/Richtlinien_zu

r_Anerkennung_als_Diabetologe_DDG_19.09.2013.pdf

DGG, & DGGG. (2018). S3-Leitlinie Gestationsdiabetes mellitus (GDM), Diagnostik,

Therapie und Nachsorge. Retrieved from https://www.deutsche-diabetes-

gesellschaft.de/fileadmin/Redakteur/Leitlinien/Evidenzbasierte_Leitlinien/2018/057-

008l_S3_Gestationsdiabetes-mellitus-GDM-Diagnostik-Therapie-Nachsorge_2018-

03.pdf

DGGG. (2010). Absolute und relative Indikationen zur Sectio cesarea. Retrieved from

https://www.dggg.de/fileadmin/documents/leitlinien/archiviert/federfuehrend/015054

_Absolute_und_relative_Indikationen_zur_Sectio_caesarea/015054_2010.pdf

Diehl, K., Schneider, S., Bock, C., Maul, H., Kleinwechter, H., & Görig, T. (2016). German

gynecologists’ experience with a universal screening for gestational diabetes mellitus

in daily practice: A qualitative study. Journal of the Turkish German Gynecological

Association, 17(1), 10–5. https://doi.org/10.5152/jtgga.2016.16182

Duran, A., Śaenz, S., Torrejon, M. J., Bordí U, E., Del Valle, L., Galindo, M., … Calle-

Pascual, A. L. (2014). Introduction of IADPSG criteria for the screening and diagnosis

of gestational diabetes mellitus results in improved pregnancy outcomes at a lower cost

in a large cohort of pregnant women: The St. Carlos gestationaldiabetes study. Diabetes

Care, 37(9), 2442–2450. https://doi.org/10.2337/dc14-0179

Eades, C. E., Cameron, D. M., & Evans, J. M. M. (2017). Prevalence of gestational diabetes

mellitus in Europe: A meta-analysis. Diabetes Research and Clinical Practice.

https://doi.org/10.1016/j.diabres.2017.03.030

Elixhauser, A., Steiner, C., Harris, D. R., & Coffey, R. M. (1998). Comorbidity measures

for use with administrative data. Medical Care, 36(1), 8–27. Retrieved from

http://www.ncbi.nlm.nih.gov/pubmed/9431328

G-BA. (2004). Anforderungen an strukturierte Behandlungsprogramme für Patienten mit

Diabetes mellitus Typ 1. Berlin. Retrieved from https://www.g-ba.de/downloads/40-

268-600/2008-05-15-DMP-Dia1-Aktualisierung.pdf

G-BA. (2012). Screening auf Gestationsdiabetes. Berlin. Retrieved from https://www.g-

71

ba.de/downloads/40-268-2557/2011-12-15_Mu-RL_Screening_GDM_ZD.pdf

G-BA. (2016). Richtlinien des Gemeinsamen Bundesausschusses über die ärztliche

Betreuung während der Schwangerschaft und nach der Entbindung („Mutterschafts-

Richtlinien“). Berlin. Retrieved from https://www.g-ba.de/downloads/62-492-

1223/Mu-RL_2016-04-21_iK-2016-07-20.pdf

Garbe, E., Suling, M., Kloss, S., Lindemann, C., & Schmid, U. (2011). Linkage of mother-

baby pairs in the German Pharmacoepidemiological Research Database.

Pharmacoepidemiology and Drug Safety, 20(3), 258–264.

https://doi.org/10.1002/pds.2038

Gherman, R. B., Chauhan, S., Ouzounian, J. G., Lerner, H., Gonik, B., & Goodwin, T. M.

(2006). Shoulder dystocia: The unpreventable obstetric emergency with empiric

management guidelines. American Journal of Obstetrics and Gynecology, 195(3), 657–

672. https://doi.org/10.1016/j.ajog.2005.09.007

Görig, T., Schneider, S., Bock, C., Maul, H., Kleinwechter, H., & Diehl, K. (2015).

Screening for gestational diabetes mellitus in Germany: A qualitative study on pregnant

women’s attitudes, experiences, and suggestions. Midwifery, 31(11), 1026–1031.

https://doi.org/10.1016/j.midw.2015.07.001

Han, S., & Crowther, C. (2013). Different types of dietary advice for women with gestational

diabetes mellitus. … Database Syst Rev, (3). Retrieved from

http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD009275.pub2/pdf/standard

Harlass, F. E., Brady, K., & Read, J. A. (1991). Reproducibility of the oral glucose tolerance

test in pregnancy. American Journal of Obstetrics and Gynecology, 164(2), 564–8.

Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/1992702

Homko, C., Sivan, E., Chen, X., Reece, E. A., & Boden, G. (2001). Insulin Secretion during

and after Pregnancy in Patients with Gestational Diabetes Mellitus 1. The Journal of

Clinical Endocrinology & Metabolism, 86(2), 568–573.

https://doi.org/10.1210/jcem.86.2.7137

Huhn, E. A., Fischer, T., Göbl, C. S., Todesco Bernasconi, M., Kreft, M., Kunze, M., …

Hoesli, I. (2016). Screening of gestational diabetes mellitus in early pregnancy by oral

glucose tolerance test and glycosylated fibronectin: study protocol for an international,

prospective, multicentre cohort trial. BMJ Open, 6(10), e012115.

72

https://doi.org/10.1136/bmjopen-2016-012115

Hunt, K. J., & Schuller, K. L. (2007). The Increasing Prevalence of Diabetes in Pregnancy.

Obstetrics and Gynecology Clinics of North America, 34(2), 173–199.

https://doi.org/10.1016/j.ogc.2007.03.002

IADPSG. (2010). International Association of Diabetes and Pregnancy Study Groups

recommendations on the diagnosis and classification of hyperglycemia in pregnancy.

Diabetes Care, 33(3), 676–682. https://doi.org/10.2337/dc09-1848

Ignell, C., Claesson, R., Anderberg, E., & Berntorp, K. (2014). Trends in the prevalence of

gestational diabetes mellitus in southern Sweden, 2003-2012. Acta Obstetricia et

Gynecologica Scandinavica, 93(4), 420–424. https://doi.org/10.1111/aogs.12340

InEK GmbH. (2016). G-DRG Fallpauschalen-Katalog 2017. Retrieved from https://www.g-

drg.de/G-DRG-System_2017/Fallpauschalen-Katalog2/Fallpauschalen-Katalog_2017

Jürges, H., & Köberlein, J. (2015). What explains DRG upcoding in neonatology? The roles

of financial incentives and infant health. Journal of Health Economics, 43, 13–26.

https://doi.org/10.1016/j.jhealeco.2015.06.001

Kassenärztliche Bundesvereinigung. (2015). Richtlinie der Kassenärztlichen

Bundesvereinigung nach § 75 Absatz 7 SGB V zur Vergabe Praxisnetznummern, 1–14.

Retrieved from http://www.kbv.de/media/sp/Arztnummern_Richtlinie.pdf

Kassenärztliche Bundesvereinigung. (2016). EBM 02311 - Behandlung Diabetischer Fuß.

Retrieved from

http://www.kbv.de/tools/ebm/html/02311_2902002740192336862528.html

Kautzky-Willer, A., Prager, R., Waldhäusl, W., Pacini, G., Thomaseth, K., Wagner, O. F.,

… Ludvik, B. (1997). Pronounced insulin resistance and inadequate β-cell secretion

characterize lean gestational diabetes during and after pregnancy. Diabetes Care,

20(11), 1717–1723. https://doi.org/10.2337/diacare.20.11.1717

Keag, O. E., Norman, J. E., & Stock, S. J. (2018). Long-term risks and benefits associated

with cesarean delivery for mother, baby, and subsequent pregnancies: Systematic

review and meta-analysis. PLoS Medicine, 15(1), 1–22.

https://doi.org/10.1371/journal.pmed.1002494

Kerstjens, J. M., Bocca-Tjeertes, I. F., de Winter, A. F., Reijneveld, S. A., & Bos, A. F.

73

(2012). Neonatal Morbidities and Developmental Delay in Moderately Preterm-Born

Children. PEDIATRICS, 130(2), e265–e272. https://doi.org/10.1542/peds.2012-0079

Kleinwechter, H., Schäfer-Graf, U., Bührer, C., Hoesli, I., Kainer, F., Kautzky-Willer, A.,

… Sorger, M. (2011). Evidenzbasierte Leitlinie zu Diagnostik, Therapie u. Nachsorge

der Deutschen Diabetes-Gesellschaft (DDG) und der Deutschen Gesellschaft für

Gynäkologie und Geburtshilfe (DGGG) Herausgeber. Diabetologie Und Stoffwechsel,

6(05), 290–328. https://doi.org/10.1055/s-00000134

Landon, M. B., Spong, C. Y., Thom, E., Carpenter, M. W., Ramin, S. M., Casey, B., …

Anderson, G. B. (2009). A Multicenter, Randomized Trial of Treatment for Mild

Gestational Diabetes. New England Journal of Medicine, 361(14), 1339–1348.

https://doi.org/10.1056/NEJMoa0902430

Lange, A., Prenzler, A., Bachmann, O., Linder, R., Neubauer, S., Zeidler, J., … von der

Schulenburg, J. M. (2015). Regional differences in health care of patients with

inflammatory bowel disease in Germany. Health Economics Review, 5(1), 1–11.

https://doi.org/10.1186/s13561-015-0067-1

MacDonald, S. C., Hernán, M. A., McElrath, T. F., & Hernández-Díaz, S. (2018).

Assessment of recording bias in pregnancy studies using health care databases: An

application to neurologic conditions. Paediatric and Perinatal Epidemiology, 32(3),

281–286. https://doi.org/10.1111/ppe.12459

Macfarlane, C. M., & Tsakalakos, N. (1988). The extended Pedersen hypothesis. Clinical

Physiology and Biochemistry, 6(2), 68–73. Retrieved from

http://www.ncbi.nlm.nih.gov/pubmed/3402161

McCloskey, L., Bernstein, J., Winter, M., Iverson, R., & Lee-Parritz, A. (2014). Follow-Up

of Gestational Diabetes Mellitus in an Urban Safety Net Hospital: Missed Opportunities

to Launch Preventive Care for Women. Journal of Women’s Health, 23(4), 327–334.

https://doi.org/10.1089/jwh.2013.4628

McIntyre, H. D., Metzger, B. E., Lowe, L. P., Catalano, P. M., Trimble, E. R., Oats, J. J. N.,

… Coustan, D. R. (2010). Hyperglycaemia and adverse pregnancy outcome (HAPO)

study: Associations with maternal body mass index. BJOG: An International Journal

of Obstetrics and Gynaecology, 117(5), 575–584. https://doi.org/10.1111/j.1471-

0528.2009.02486.x

74

Melchior, H., Kurch-Bek, D., & Mund, M. (2017). The prevalence of gestational diabetes.

Deutsches Arzteblatt International, 114(24), 412–418.

https://doi.org/10.3238/arztebl.2017.0412

Metcalfe, A., Lix, L. M., Johnson, J. A., Currie, G., Lyon, A. W., Bernier, F., & Tough, S.

C. (2015). Validation of an obstetric comorbidity index in an external population.

BJOG: An International Journal of Obstetrics and Gynaecology, 122(13), 1748–1755.

https://doi.org/10.1111/1471-0528.13254

Metzger, B. E., & Coustan, D. R. (1998). Summary and recommendations of the Fourth

International Workshop-Conference on Gestational Diabetes Mellitus. The Organizing

Committee. Diabetes Care, 21 Suppl 2, B161-7. Retrieved from

http://www.ncbi.nlm.nih.gov/pubmed/9704245

Metzger, B. E., Lowe, L. P., Dyer, A. R., Trimble, E. R., Chaovarindr, U., Coustan, D. R.,

… Sacks, D. A. (2008). Hyperglycemia and adverse pregnancy outcomes. The New

England Journal of Medicine, 358(19), 1991–2002.

https://doi.org/10.1056/NEJMoa0707943

Mikolajczyk, R., Kraut, A., & Garbe, E. (2013). Evaluation of pregnancy outcome records

in the German Pharmacoepidemiological Research Database (GePaRD).

Pharmacoepidemiology and Drug Safety, 22(8), 873–880.

https://doi.org/10.1002/pds.3467

Naylor, C. D., Sermer, M., Chen, E., & Sykora, K. (1996). Cesarean delivery in relation to

birth weight and gestational glucose tolerance: pathophysiology or practice style?

Toronto Trihospital Gestational Diabetes Investigators. JAMA, 275(15), 1165–70.

https://doi.org/10.1001/jama.1996.03530390031030

NICE. (2015). Diabetes in Pregnancy: Management of Diabetes and Its Complications from

Preconception to the Postnatal Period. Diabetes in Pregnancy: Management of

Diabetes and Its Complications from Preconception to the Postnatal Period,

2008(July), e619–e626. https://doi.org/25950069

Ohlmeier, C., Frick, J., Prütz, F., Lampert, T., Ziese, T., Mikolajczyk, R., & Garbe, E.

(2014). Nutzungsmöglichkeiten von Routinedaten der Gesetzlichen

Krankenversicherung in der Gesundheitsberichterstattung des Bundes.

Bundesgesundheitsblatt - Gesundheitsforschung - Gesundheitsschutz, 57(4), 464–472.

75

https://doi.org/10.1007/s00103-013-1912-1

Robert Koch-Institut. (2018). AdiMon-Themenblatt: Geburtsgewicht (Stand 26. Februar

2018).

Rozance, P. J., & Hay, W. W. (2010). Describing hypoglycemia - Definition or operational

threshold? Early Human Development, 86(5), 275–280.

https://doi.org/10.1016/j.earlhumdev.2010.05.002

Ryser Rüetschi, J., Jornayvaz, F., Rivest, R., Huhn, E., Irion, O., & Boulvain, M. (2016).

Fasting glycaemia to simplify screening for gestational diabetes. BJOG: An

International Journal of Obstetrics and Gynaecology, 2219–2222.

https://doi.org/10.1111/1471-0528.13857

Schaefer-Graf, U. M., Graf, K., Kulbacka, I., Kjos, S. L., Dudenhausen, J., Vetter, K., &

Herrera, E. (2008). Maternal lipids as strong determinants of fetal environment and

growth in pregnancies with gestational diabetes mellitus. Diabetes Care, 31(9), 1858–

1863. https://doi.org/10.1016/j.ajog.2010.11.039.

Schneeweiss, S., & Avorn, J. (2005). A review of uses of health care utilization databases

for epidemiologic research on therapeutics. Journal of Clinical Epidemiology, 58(4),

323–337. https://doi.org/10.1016/j.jclinepi.2004.10.012

Schneeweiss, S., & Maclure, M. (2000). Use of Comorbidity Scores for Control of

Confounding in Studies Using Administrative Databases. International Journal of

Epidemiology, 29(5), 891–898. https://doi.org/10.1093/ije/29.5.891

Schneider, U., Linder, R., & Verheyen, F. (2016). Long-term sick leave and graded return

to work: What do we know about the follow-up effects? Health Policy, 120(10), 1193–

1201. https://doi.org/10.1016/j.healthpol.2016.09.001

Schreyögg, J., & Stargardt, T. (2012). Gesundheitsökonomische Evaluation auf Grundlage

von GKV-Routinedaten. Bundesgesundheitsblatt - Gesundheitsforschung -

Gesundheitsschutz, 55(5), 668–676. https://doi.org/10.1007/s00103-012-1476-5

Sibai, B., Dekker, G., & Kupferminc, M. (2005). Pre-eclampsia. The Lancet, 365(9461),

785–799. https://doi.org/10.1016/S0140-6736(05)17987-2

Sivan, E., Homko, C. J., Chen, X., Reece, E. A., & Boden, G. (1999). Effect of insulin on

fat metabolism during and after normal pregnancy. Diabetes, 48(4), 834–838.

76

https://doi.org/10.2337/diabetes.48.4.834

Song, C., Lyu, Y., Li, C., Liu, P., Li, J., Ma, R. C., & Yang, X. (2018). Long-term risk of

diabetes in women at varying durations after gestational diabetes: a systematic review

and meta-analysis with more than 2 million women. Obesity Reviews, 19(3), 421–429.

https://doi.org/10.1111/obr.12645

Tamayo, T., Tamayo, M., Rathmann, W., & Potthoff, P. (2016). Prevalence of gestational

diabetes and risk of complications before and after initiation of a general systematic

two-step screening strategy in Germany (2012-2014). Diabetes Research and Clinical

Practice. https://doi.org/10.1016/j.diabres.2016.03.001

Van Walraven, C., Austin, P. C., Jennings, A., Quan, H., & Forster, A. J. (2009). A

modification of the elixhauser comorbidity measures into a point system for hospital

death using administrative data. Medical Care, 47(6), 626–633.

https://doi.org/10.1097/MLR.0b013e31819432e5

Zhang, C., & Ning, Y. (2011). Effect of dietary and lifestyle factors on the risk of gestational

diabetes: review of epidemiologic evidence. The American Journal of Clinical

Nutrition, 94(suppl_6), 1975S–1979S. https://doi.org/10.3945/ajcn.110.001032

Zou, K. H., O’Malley, A. J., & Mauri, L. (2007). Receiver-operating characteristic analysis

for evaluating diagnostic tests and predictive models. Circulation, 115(5), 654–657.

https://doi.org/10.1161/CIRCULATIONAHA.105.594929

77

10. Appendix

The appendix is split into two separate section. The first section includes additional

information, including ICD-10 and OPS codes, on the method section. The second section

includes full logistic regressions of all outcome parameters.

10.1. Methods

Table 29 ICD-10 codes used for identification of pregnancies

ICD-CODE DESCRIPTION

O151 Eclampsia in labour

O152 Eclampsia in the puerperium

O48 Postdate pregnancy

O601 Preterm labour with preterm delivery

O602 Preterm labour with delivery

O603 Preterm delivery without spontaneous labour

O61 Failed induction

O62 Abnormalities of forces of labour

O63 Long labour

O64 Obstructed labour due to malposition and malpresentation of fetus

O65 Obstructed labour due to maternal pelvic abnormality

O66 Other obstructed labour

O67 Labour and delivery complicated by intrapartum haemorrhage, not elsewhere classified

O68 Labour and delivery complicated by foetal stress

O69 Labour and delivery complicated by umbilical cord complications

O70 Perineal laceration during delivery

O71 Other obstetric trauma

O72 Postpartum haemorrhage

O73 Retained placenta and membranes, without haemorrhage

O74 Complications of anaesthesia during labour and delivery

O75 Other complications of labour and delivery, not elsewhere classified

O80 Single spontaneous delivery

O81 Single delivery by forceps and vacuum extractor

O82 Single delivery by caesarean section

O85 Puerperal sepsis

O86 Other puerperal infections

O87 Venous complications in the puerperium

O88 Obstetric embolism

78

Tabelle 30 continued

O89 Complications of anaesthesia during the puerperium

O90 Complications of the puerperium, not elsewhere classified

P05 Slow foetal growth (in newborn)

P07 Short gestation and low birth weight

P08 Long gestation and high birth weight

Z370! Single live birth

Z371! Single stillbirth

Z372! Twins, both liveborn

Z373! Twins, one liveborn and one stillborn

Z374! Twins, both stillborn

Z375! Other multiple births, all liveborn

Z376! Other multiple births, some liveborn

Z377! Other multiple births, all stillborn

Z38 Liveborn infants according to place of birth

Z39 Postpartum care and examination

P95 Stillbirth

P072 Extreme immaturity (less than 28 completed weeks of gestation).

P073 Other preterm infants 28 completed weeks or more but less than 37 completed weeks of

gestation.

Table 30 OPS-codes used to identify pregnancies

OPS DESCRIPTION

572 Birth with breech presentation and instrumental delivery

5730 Artificial Amnotomie

5731 Other excised induction of labor

5732 Internal and combined version with and without extraction

5733 Failed vaginal excised induction of labor

5738 Episiotomie

5739 Other operations for induction of labour

5740 Classical sectio cesarea

5741 Sectio cesarea supracervical and corporal

5742 Sectio cesarea extraperitonealis

5745 Sectio cesarea with other gynaecological intervention

5749 Other sectio cesarea

5756 Ablation of the remaining placenta

5758 Reconstruction of female genitals after rupture

926 Birth accompanying measures

9280 Inpatient treatment before birth at the same hospitalisation

79

Table 31 ICD-10 codes defining comorbidities within the OCI

Condition Codes

Alcohol abuse F10

Asthma J44, J45

Cardiac Valvular Disease I05-I09, I34-I39

Chronic Congestive Heart Failure I50.0

Chronic Ischemic Heart Disease I20, I25

Chronic Renal Disease N02.2, N03 - N05, N08,

N17.1, N17.2, N18, N25, O26.8

Congenital Heart Disease Q20 - Q26, O99.4

Drug Abuse F11 - F16, F18, F19

Gestational hypertension O13, O16

Human Immunodeficiency Virus B20, B24, O98.7, Z21

Mild/Unspecified Pre-Eclampsia O11, O14

Placenta Previa O44

Pre-Exisiting Diabetes Mellitus E10, E11, O24.5 – O24.7

Pre-Existing Hypertension I10 - I13, I15, O10, O11

Previous Cesarean Delivery O34.20

Pulmonary Hypertension I27.0, I27.2, I27.8, I27.9

Severe Pre-Eclampsia O14, O15

Sickle Cell Disease D56, D57

Systemic Lupus Erythematosus M32

80

Table 32 ICD-10 codes defining end-organ damage

Condition Codes

Acute Heart Failure I26.0, I46, I50, I97.8,

I97.9, O75.4

Acute Liver Disease K72.0, K72.9, O26.6

Acute Myocardial Infarction I21, I22

Acute Renal Failure N17, O90.4

Acute Respiratory Distress J80, J95.1, J95.2, J95.3

Syndrome/Respiratory Failure J95.8, J95.9, J96.0, J96.9,

R09.2

Coma E10.0, E11.0, E15, K72.9,

R40.2

Delirium F05, F06.0, F06.1, F06.2,

F06.3, F06.4, F06.8

Disseminated Intravascular

Coagulation/Coagulopathy

D65, D68.3, D68.4, D68.8,

D68.9, D69.5, O72.3

Puerperal Cerebrovascular Disorders G08, G43, G93.1, G93.4,

G93.6, G97.8, G97.9, I60

I63, I67.4, I67.6, I67.8,

I97.8, I97.9, O22.5, O22.8,

O22.9, O87.3, O99.4

Pulmonary Edema I50.1, J81

Pulmonary Embolism I26, O88

Sepsis A40, A41, B37.7, O75.3,

R57.2, R65.1

Shock O75.1, R57, R65.1, T78.0,

T78.2, T80.5, T81.1,

Status Asthmaticus J45.01, J45.11, J45.81, J45.91

Status Epilepticus G41

81

10.2. Results

Table 33 Logistic regression Group B: Birth weight >90th percentile

Parameter Estimate Standard

Error

Wald

Chi-Square

Pr > ChiSq Odds Ratio 95% Wald

Confidence Limits

Intercept -4.0960 0.1915 457.5331 <.0001

Group B 0.1223 0.0972 1.5839 0.2082 1.130 0.934 1.367

Age 25 – 29 0.2265 0.2133 1.1271 0.2884 1.254 0.826 1.905

Age 30 – 34 -0.0608 0.1690 0.1293 0.7191 0.941 0.676 1.311

Age 35 – 39 0.00977 0.1595 0.0038 0.9511 1.010 0.739 1.380

Age 40+ 0.0542 0.1640 0.1090 0.7413 1.056 0.765 1.456

Female

neonate

-0.6593 0.0725 82.6206 <.0001 0.517 0.449 0.596

Multiple

gestation

-2.7603 0.7075 15.2202 <.0001 0.063 0.016 0.253

Obesity,

Grade 1

0.7478 0.1654 20.4271 <.0001 2.112 1.527 2.921

Obesity,

Grade 2

0.4469 0.2195 4.1460 0.0417 1.563 1.017 2.404

Obesity,

Grade 3

1.1910 0.1919 38.5221 <.0001 3.290 2.259 4.793

R² 0.0032

Max-rescaled

R²

0.0224

Number of

observations

57,897

82

Table 34 Logistic regression Group B: Primary cesearan sectio

Parameter Estimate Standard

Error

Wald

Chi-Square

Pr > ChiSq Odds Ratio 95% Wald

Confidence Limits

Intercept -1.9968 0.0692 832.6512 <.0001

Group B 0.1493 0.0328 20.6774 <.0001 1.161 1.089 1.238

Age 25 – 29 -0.0496 0.0645 0.5906 0.4422 0.952 0.839 1.080

Age 30 – 34 0.0275 0.0616 0.1988 0.6557 1.028 0.911 1.160

Age 35 – 39 0.3009 0.0625 23.1438 <.0001 1.351 1.195 1.527

Age 40+ 0.7062 0.0721 95.8847 <.0001 2.026 1.759 2.334

Multiple

gestation

1.0834 0.0666 264.3766 <.0001 2.955 2.593 3.367

Obesity, Grade 1 0.4030 0.0674 35.7545 <.0001 1.496 1.311 1.708

Obesity, Grade 2 0.5687 0.0756 56.5757 <.0001 1.766 1.523 2.048

Obesity, Grade 3 0.7605 0.0885 73.8916 <.0001 2.139 1.799 2.544

R² 0.0133

Max-rescaled R² 0.0225

Number of

observations

56,890

83

Table 35 Logistic regression Group B: Neonatal hypoglycaemia

Parameter Estimate Standard

Error

Wald

Chi-Square

Pr > ChiSq Odds Ratio 95% Wald

Confidence Limits

Intercept -4.7056 0.1581 886.3242 <.0001

Group B 0.8682 0.0644 181.9839 <.0001 2.383 2.100 2.703

Age 25 – 29 -0.2518 0.1509 2.7826 0.0953 0.777 0.578 1.045

Age 30 – 34 -0.0852 0.1425 0.3574 0.5499 0.918 0.695 1.214

Age 35 – 39 -0.1577 0.1465 1.1591 0.2816 0.854 0.641 1.138

Age 40+ 0.0838 0.1706 0.2411 0.6234 1.087 0.778 1.519

Female

neonate

-0.2332 0.0561 17.3039 <.0001 0.792 0.710 0.884

Multiple

gestation

1.7067 0.0796 459.6386 <.0001 5.511 4.715 6.441

Obesity,

Grade 1

0.4165 0.1466 8.0689 0.0045 1.517 1.138 2.022

Obesity,

Grade 2

0.4034 0.1710 5.5661 0.0183 1.497 1.071 2.093

Obesity,

Grade 3

0.6335 0.1890 11.2421 0.0008 1.884 1.301 2.729

R² 0.0098

Max-rescaled

R²

0.0493

Number of

observations

57,897

84

Table 36 Logistic regression Group B: Premature delivery

Parameter Estimate Standard

Error

Wald

Chi-Square

Pr > ChiSq Odds Ratio 95% Wald

Confidence Limits

Intercept -3.1016 0.1136 745.1151 <.0001

Group B 0.1438 0.0549 6.8544 0.0088 1.155 1.037 1.286

Age 25 – 29 -0.1071 0.1035 1.0713 0.3007 0.898 0.734 1.100

Age 30 – 34 -0.1609 0.0992 2.6297 0.1049 0.851 0.701 1.034

Age 35 – 39 -0.1851 0.1019 3.3009 0.0692 0.831 0.681 1.015

Age 40+ 0.0468 0.1201 0.1517 0.6970 1.048 0.828 1.326

Female

neonate

-0.1893 0.0388 23.8276 <.0001 0.828 0.767 0.893

Multiple

gestation

2.8425 0.0502 3200.8176 <.0001 17.159 15.550 18.935

Obesity,

Grade 1

0.1694 0.1183 2.0494 0.1523 1.185 0.939 1.494

Obesity,

Grade 2

0.1604 0.1386 1.3392 0.2472 1.174 0.895 1.541

Obesity,

Grade 3

0.2924 0.1614 3.2827 0.0700 1.340 0.976 1.838

R² 0.0440

Max-rescaled

R²

0.1278

Number of

observations

57,897

85

Table 37 Logistic regression Group B: Shoulder dystocia

Parameter Estimate Standard

Error

Wald

Chi-Square

Pr > ChiSq Odds Ratio 95% Wald

Confidence Limits

Intercept -4.9986 0.3396 216.6486 <.0001

Group B -0.0769 0.1729 0.1975 0.6567 0.926 0.660 1.300

Age 25 – 29 0.00263 0.3040 0.0001 0.9931 1.003 0.553 1.819

Age 30 – 34 0.1124 0.2904 0.1499 0.6986 1.119 0.633 1.977

Age 35 – 39 -0.1223 0.3023 0.1636 0.6859 0.885 0.489 1.600

Age 40+ -0.1006 0.3805 0.0699 0.7915 0.904 0.429 1.906

Female

neonate

-0.3599 0.1151 9.7815 0.0018 0.698 0.557 0.874

Multiple

gestation

-0.7956 0.4517 3.1023 0.0782 0.451 0.186 1.094

Obesity,

Grade 1

0.7740 0.2673 8.3867 0.0038 2.168 1.284 3.661

Obesity,

Grade 2

0.4476 0.3607 1.5397 0.2147 1.565 0.772 3.173

Obesity,

Grade 3

0.5744 0.4156 1.9101 0.1670 1.776 0.786 4.011

R² 0.0005

Max-rescaled

R²

0.0070

Number of

observations

57,897

86

Table 38 Logistic regression Group B: Intensive Neonatal care

Parameter Estimate Standard

Error

Wald

Chi-Square

Pr > ChiSq Odds Ratio 95% Wald

Confidence Limits

Intercept -2.0947 0.0786 709.7764 <.0001

Group B 0.1531 0.0399 14.7575 0.0001 1.165 1.078 1.260

Age 25 – 29 -0.2277 0.0707 10.3616 0.0013 0.796 0.693 0.915

Age 30 – 34 -0.2589 0.0675 14.7089 0.0001 0.772 0.676 0.881

Age 35 – 39 -0.3075 0.0697 19.4667 <.0001 0.735 0.641 0.843

Age 40+ -0.1316 0.0853 2.3806 0.1228 0.877 0.742 1.036

Female

neonate

-0.3010 0.0281 114.6312 <.0001 0.740 0.700 0.782

Multiple

gestation

1.9385 0.0479 1636.2895 <.0001 6.948 6.326 7.633

Obesity,

Grade 1

0.3250 0.0820 15.7147 <.0001 1.384 1.179 1.625

Obesity,

Grade 2

0.2757 0.0970 8.0782 0.0045 1.317 1.089 1.593

Obesity,

Grade 3

0.4972 0.1100 20.4309 <.0001 1.644 1.325 2.040

R² 0.0265

Max-rescaled

R²

0.0546

Number of

observations

57,897

87

Table 39 Logistic regression Group B: neonatal hyperbilirubinemia

Parameter Estimate Standard

Error

Wald

Chi-Square

Pr > ChiSq Odds Ratio 95% Wald

Confidence Limits

Intercept -3.2047 0.1177 740.7644 <.0001

Group B 0.1990 0.0537 13.7558 0.0002 1.220 1.098 1.356

Age 25 – 29 0.0879 0.1085 0.6561 0.4179 1.092 0.883 1.350

Age 30 – 34 0.0686 0.1044 0.4320 0.5110 1.071 0.873 1.314

Age 35 – 39 0.0281 0.1071 0.0689 0.7930 1.029 0.834 1.269

Age 40+ 0.2881 0.1243 5.3765 0.0204 1.334 1.046 1.702

Female

neonate

-0.2913 0.0388 56.4983 <.0001 0.747 0.693 0.806

Multiple

gestation

1.3178 0.0652 408.1835 <.0001 3.735 3.287 4.245

Obesity,

Grade 1

0.1875 0.1160 2.6121 0.1061 1.206 0.961 1.514

Obesity,

Grade 2

0.1049 0.1401 0.5608 0.4539 1.111 0.844 1.462

Obesity,

Grade 3

0.3615 0.1539 5.5153 0.0189 1.435 1.062 1.941

R² 0.0071

Max-rescaled

R²

0.0216

Number of

observations

57,897

88

Table 40 Logistic regression Group B: Preeclampsia

Parameter Estimate Standard

Error

Wald

Chi-Square

Pr > ChiSq Odds Ratio 95% Wald

Confidence Limits

Intercept -3.0467 0.0887 1180.5538 <.0001

Group B 0.2808 0.0426 43.4259 <.0001 1.324 1.218 1.440

Age 25 – 29 -0.0897 0.0823 1.1876 0.2758 0.914 0.778 1.074

Age 30 – 34 -0.1573 0.0788 3.9823 0.0460 0.854 0.732 0.997

Age 35 – 39 -0.1123 0.0806 1.9408 0.1636 0.894 0.763 1.047

Age 40+ 0.0825 0.0961 0.7379 0.3903 1.086 0.900 1.311

Multiple

gestation

0.8628 0.0828 108.4439 <.0001 2.370 2.015 2.788

Obesity, Grade 1 0.6402 0.0823 60.4723 <.0001 1.897 1.614 2.229

Obesity, Grade 2 1.0357 0.0833 154.6594 <.0001 2.817 2.393 3.317

Obesity, Grade 3 1.3578 0.0926 214.9912 <.0001 3.888 3.242 4.661

R² 0.0133

Max-rescaled R² 0.0225

Number of

observations

56,890

89

Table 41 Logistic regression Group C: Premature delivery

Parameter Estimate Standard

Error

Wald

Chi-Square

Pr > ChiSq Odds Ratio 95% Wald

Confidence Limits

Intercept -3.1169 0.1688 341.0857 <.0001

Group C 0.1984 0.1352 2.1549 0.1421 1.219 0.936 1.589

Age 25 – 29 -0.1518 0.1075 1.9926 0.1581 0.859 0.696 1.061

Age 30 – 34 -0.2002 0.1029 3.7836 0.0518 0.819 0.669 1.002

Age 35 – 39 -0.2713 0.1063 6.5128 0.0107 0.762 0.619 0.939

Age 40+ 0.0137 0.1273 0.0116 0.9143 1.014 0.790 1.301

Female

neonate

-0.1892 0.0416 20.7162 <.0001 0.828 0.763 0.898

Multiple

gestation

2.8949 0.0542 2851.8829 <.0001 18.081 16.259 20.108

Obesity,

Grade 1

0.1833 0.1324 1.9187 0.1660 1.201 0.927 1.557

Obesity,

Grade 2

0.2963 0.1472 4.0511 0.0441 1.345 1.008 1.795

Obesity,

Grade 3

0.1901 0.1834 1.0747 0.2999 1.209 0.844 1.732

R² 0.0435

Max-rescaled

R²

0.1283

Number of

observations

51,918

90

Table 42 Logistic regression Group C: Shoulder dystocia

Parameter Estimate Standard

Error

Wald

Chi-Square

Pr > ChiSq Odds Ratio 95% Wald

Confidence Limits

Intercept -5.2539 0.4774 121.1391 <.0001

Group C 0.1144 0.3665 0.0975 0.7549 1.121 0.547 2.300

Age 25 – 29 0.0427 0.3294 0.0168 0.8969 1.044 0.547 1.991

Age 30 – 34 0.1817 0.3150 0.3329 0.5640 1.199 0.647 2.223

Age 35 – 39 0.0440 0.3254 0.0183 0.8925 1.045 0.552 1.978

Age 40+ -0.1982 0.4280 0.2144 0.6434 0.820 0.354 1.898

Female

neonate

-0.4167 0.1218 11.6981 0.0006 0.659 0.519 0.837

Multiple

gestation

-0.6532 0.4522 2.0865 0.1486 0.520 0.215 1.262

Obesity,

Grade 1

0.8973 0.2772 10.4760 0.0012 2.453 1.425 4.223

Obesity,

Grade 2

0.4531 0.3872 1.3692 0.2419 1.573 0.737 3.360

Obesity,

Grade 3

0.7204 0.4190 2.9563 0.0855 2.055 0.904 4.672

R² 0.0006

Max-rescaled

R²

0.0084

Number of

observations

51,918

91

Table 43 Logistic regression Group C: Intensive neonatal Care

Parameter Estimate Standard

Error

Wald

Chi-Square

Pr > ChiSq Odds Ratio 95% Wald

Confidence Limits

Intercept -2.1393 0.1169 335.0545 <.0001

Group C 0.2231 0.0945 5.5689 0.0183 1.250 1.039 1.504

Age 25 – 29 -0.2583 0.0743 12.0989 0.0005 0.772 0.668 0.893

Age 30 – 34 -0.2864 0.0708 16.3489 <.0001 0.751 0.654 0.863

Age 35 – 39 -0.3321 0.0733 20.5170 <.0001 0.717 0.621 0.828

Age 40+ -0.1099 0.0905 1.4730 0.2249 0.896 0.750 1.070

Female

neonate

-0.3163 0.0301 110.7917 <.0001 0.729 0.687 0.773

Multiple

gestation

1.9735 0.0518 1454.0911 <.0001 7.196 6.502 7.964

Obesity,

Grade 1

0.3227 0.0918 12.3540 0.0004 1.381 1.153 1.653

Obesity,

Grade 2

0.3185 0.1054 9.1235 0.0025 1.375 1.118 1.691

Obesity,

Grade 3

0.5423 0.1190 20.7853 <.0001 1.720 1.362 2.172

R² 0.0264

Max-rescaled

R²

0.0549

Number of

observations

51,918

92

Table 44 Logistic regression Group C: Hyperbilirubinemia

Parameter Estimate Standard

Error

Wald

Chi-Square

Pr > ChiSq Odds Ratio 95% Wald

Confidence Limits

Intercept -2.7795 0.1912 211.3335 <.0001

Group C -0.2319 0.1572 2.1753 0.1402 0.793 0.583 1.079

Age 25 – 29 0.0495 0.1156 0.1830 0.6688 1.051 0.838 1.318

Age 30 – 34 0.0799 0.1110 0.5187 0.4714 1.083 0.871 1.346

Age 35 – 39 0.0305 0.1141 0.0717 0.7889 1.031 0.824 1.289

Age 40+ 0.3200 0.1334 5.7553 0.0164 1.377 1.060 1.789

Female

neonate

-0.2920 0.0418 48.9075 <.0001 0.747 0.688 0.810

Multiple

gestation

1.3962 0.0698 400.3999 <.0001 4.040 3.523 4.632

Obesity,

Grade 1

0.2003 0.1317 2.3112 0.1284 1.222 0.944 1.582

Obesity,

Grade 2

0.0824 0.1611 0.2616 0.6090 1.086 0.792 1.489

Obesity,

Grade 3

0.3046 0.1782 2.9218 0.0874 1.356 0.956 1.923

R² 0.0072

Max-rescaled

R²

0.0225

Number of

observations

51,918

93

Table 45 Logistic regression Group C: Preeclampsia

Parameter Estimate Standard

Error

Wald

Chi-Square

Pr > ChiSq Odds Ratio 95% Wald

Confidence Limits

Intercept -3.1930 0.1357 553.9896 <.0001

Group B 0.4126 0.1030 16.0353 <.0001 1.511 1.235 1.849

Age 25 – 29 -0.0689 0.0969 0.5059 0.4769 0.933 0.772 1.129

Age 30 – 34 -0.1862 0.0934 3.9773 0.0461 0.830 0.691 0.997

Age 35 – 39 -0.0893 0.0958 0.8688 0.3513 0.915 0.758 1.103

Age 40+ 0.1072 0.1165 0.8463 0.3576 1.113 0.886 1.399

Multiple

gestation

0.8734 0.1046 69.7029 <.0001 2.395 1.951 2.940

Obesity, Grade 1 0.8269 0.0961 73.9696 <.0001 2.286 1.894 2.760

Obesity, Grade 2 1.1029 0.0997 122.4601 <.0001 3.013 2.478 3.663

Obesity, Grade 3 1.2685 0.1149 121.8045 <.0001 3.556 2.838 4.454

R² 0.0070

Max-rescaled R² 0.0194

Number of

observations

51,052

94

Declaration of Independent Work

I hereby declare that I have authored this thesis independently, that I have not used other

than the declared sources/resources, and that I have explicitly marked all material which

has been quoted either literally or by content from the used sources.

Hamburg, 25.09.2018

Patrick Reinders